Academically oriented

Full-time study

4 years

prof. Ing. Lukáš Sekanina, Ph.D.

Chairman :
prof. Ing. Lukáš Sekanina, Ph.D.
Councillor internal :
prof. Dr. Ing. Pavel Zemčík, dr. h. c.
prof. Dr. Ing. Zbyněk Raida
prof. RNDr. Josef Šlapal, CSc.
prof. Ing. Pavel Václavek, Ph.D.
prof. Ing. Tomáš Hruška, CSc.
prof. RNDr. Alexandr Meduna, CSc.
doc. Dr. Ing. Petr Hanáček
prof. Dr. Ing. Jan Černocký
prof. Ing. Adam Herout, Ph.D.
doc. Ing. Jan Kořenek, Ph.D.
prof. Ing. Tomáš Vojnar, Ph.D.
Councillor external :
prof.,RNDr. Jiří Barnat, Ph.D.

The goal of the doctoral degree programme is to provide outstanding graduates from the master degree programme with a specialised university education of the highest level in certain fields of computer science and information technology, including especially the areas of information systems, computer-based systems and computer networks, computer graphics and multimedia, and intelligent systems. The education obtained within this degree programme also comprises a training and attestation for scientific work.

• Graduates from the doctoral study programme are trained to independently work in research, development, or management.
• They are able to solve and/or to lead teams solving advanced conceptual, research, development, or production problems in the area of contemporary information technology and its applications.
• They can be engaged to work on creative tasks, to lead research and development teams, or to work in management of companies or organizations whenever there are required abilities to work in an independent and creative way, to analyze complex problems, and to propose and realize new and original solutions. Graduates from the doctoral study programme can also teach and/or scientifically work at universities.

FIT graduates in general and FIT doctoral graduates in particular do not have a problem finding employment at scientific, pedagogical or management positions both in Czech Republic and abroad.

• FIT   graduates of the doctoral study are capable of independent scientific, research and management work in the field of Informatics, Computer Technology and Information Technologies. Graduates are ready to solve challenging conceptual, research and development problems. They can independently conduct research, development and production in the field of modern information technology.
• Typically, they work as creative workers at top scientific research workplaces, as leaders of research and development teams and in scientific and pedagogical work at universities. Graduates of this program are also employed in higher functional positions of larger institutions and companies, where the ability to work independently, analyze complex problems and design and implement new, original solutions is required.
• And, last but not least, graduates typically continue as so-called "postdoc" in their academic careers in Czech Republic or abroad.

The requirements that the doctoral students have to fulfil are given by their individual study plans, which specify the courses that they have to complete, their presupposed study visits and active participation at scientific conferences, and their minimum pedagogical activities within the bachelor and master degree programmes of the faculty. A successful completion of the doctoral studies is conditional on the following:

• The student has to pass a doctoral state examination within which he/she has to prove a deep knowledge of methodologies, theories, and their applications in accordance with the state of the art in the areas of science that are given by the courses included in his/her individual study plan and by the theme of his/her future dissertation thesis. The doctoral state examination also encompasses an evaluation of the presumed goals of the future dissertation thesis of the student, of the chosen solution method, and of the so far obtained original results.
• The student has further to prepare and defend his dissertation thesis.

The rules are determined by the directions of the dean for preparing the individual study plan of a doctoral student.  The plan is to be based on the theme of his/her future dissertation thesis and it is to be approved by the board of the branch.

• obligatory doctoral study programme Courses, the total number of courses a student has to complete and their mapping into particular semesters.
• a Research Plan Content (brief descrition of research content - focuse at the intended research area and the doctoral thesis topic
• a Research Plan ( list of research activities focused at the intended research area and the doctoral thesis topic - conferences and seminars to be attended , work to be published)
• teaching duty according to BUT study rules and regulations
• doctoral study schedule

Brno university of technology provides studies for persons with health disabilities according to section 21 par. 1 e) of the Act no. 111/1998, about universities and about the change and supplementing other laws (Higher Education Act) as amended, and according to the requirements in this field arising from Government Regulation No. 274/2016 Coll., on standards for accreditation in higher education, provides services for study applicants and students with specific needs within the scope and in form corresponding with the specification stated in Annex III to Rules for allocation of a financial contribution and funding for public universities by the Ministry of Education, Youth and Sports, specifying financing additional costs of studies for students with specific needs.

Services for students with specific needs at BUT are carried out through the activities of specialized workplace - Alfons counselling center, which is a part of BUT Lifelong Learning Institute - Student counselling section.

Counselling center activities and rules for making studies accessible are guaranteed by the university through a valid Rector's directive 11/2017 concerning the status of study applicants and students with specific needs at BUT. This internal standard guarantees minimal stadards of provided services.
Services of the counselling center are offered to all study applicants and students with any and all types of health disabilities stated in the Methodological standard of the Ministry of Education, Youth and Sports.

The study programme builds on both the ongoing follow-up Master's programme in Information Technology and the new follow-up Master's programme in Information Technology and Artificial Intelligence.
Students can also, according to their needs and outside their formalized studies, take courses and trainings related to the methodology of scientific work, publishing and citation skills, ethics, pedagogy and soft skills organized by BUT or other institutions.

3. round (applications submitted from 21.09.2026 to 31.01.2027)

1. Advanced algorithms of Video, Image, and/or Signal processing

   The topic concerns algorithms of image, video, and/or signal processing. Its main goal is to research and in-depth analyze existing algorithms and discover new ones so that they have desirable features and so that they are possible to efficiently implement. Such efficient implementation can be but does not necessarily have to be part of the work but it is important to prepare the algorithms so that they can be efficiently implemented e.g. in CPU, in GPU, in AI accelratorrs, embeded systems, even in combination with FPGA, in extremely low power systems, or in other environments. It is possible to exploit algorithms of artificial intelligence, such as neural networks, especially CNNs, LLM/VLM. The application possibilities of the algorithms are also important and the application can be but does not have to be part of the work. The algorithms/applications of interest include:

   • recognition of scene contents, events, and general semantics of video sequences (such as identification of traffic situations, identification in scenes in moview, action identification, etc.),
   • classification of video sequences using machine learning (AI) through deep convolution networks neural network or similar approaches (e.g. for industrial quality inspection, object of scene characteristics search, etc.), possibly in combination with object tracking in video using modern methods,
   • parallel analysis of video and signal (e.g. for detection of coincidence of occurrence of object in video and characteristic signal shape in surveillance applications), fusion of video and sognals,
   • modern algorithms of video, image, and/or signal exploiting "client/server" or "cloud" approaches suitable e.g. for mobile technology and/or embedded systems,
   • algorithms of video compression and analysis through frequency or wavelet transforms or similar methods...

   After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

   Collaboration on grant projects, such as TACR, MPO, H2020, ECSEL (employment expected).

   Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

2. Agentic AI in Chip Design and Other Parts of Electronic Design Automation

   Research in Agentic AI for chip design and Electronic Design Automation (EDA) is rapidly shifting from passive, single-point tool optimisation to autonomous, goal-driven workflows. Agentic AI systems use reasoning agents to plan, orchestrate EDA tools, learn from outcomes, and collaborate to manage multi-step tasks throughout the semiconductor design lifecycle. 

   The goal of the research is to bring AI-native autonomy to various stages of the frontend and backend components of the IC design, including RTL, verification and physical design. An initial step will be to adapt and use specialised models trained on Verilog, netlists, and layout data to understand the semantic structure of electronic circuits, translate requirements into synthesizable code, and iterate automatically using compiler feedback in the context of the RISC-V Automotive Processor explored in the Chips JU project Rigoletto.

   Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

3. Analysis of programs with complex data structure

   Software projects that employ dynamic data structures in combination with pointers are highly prone to memory-related errors (such as null pointer dereferencing, memory leaks, etc.). At the same time, such code is often system-level code used in operating system kernels, shared libraries, or interpreters of higher-level programming languages.

   The aim of this dissertation is to build upon existing techniques for the analysis of programs with dynamic memory, in particular those based on Separation Logic and bi-abduction. The work may focus on the C/C++ programming language or another suitable language. It will build on prior work by members of the VeriFIT research group and by Dr. F. Zuleger from TU Wien.

   An alternative direction of the dissertation is to build upon current techniques for automated complexity analysis, with the goal of analyzing open code and/or code with complex data structures. This direction will also build on the work of the VeriFIT group, as well as Dr. F. Zuleger from TU Wien and Dr. M. Sinn from the University of Applied Sciences St. Pölten.

   The dissertation will be carried out in collaboration with the VeriFIT team at the Faculty of Information Technology, Brno University of Technology, which focuses on program verification techniques for programs with complex data structures (in particular Dr. L. Holík, Prof. T. Vojnar, Ing. T. Dacík, Ing. V. Šoková). With a responsible approach and high-quality results, there is an opportunity to participate in research grant projects, including international ones. There is also the possibility of close collaboration with various international partners of VeriFIT: TU Wien, Austria (Dr. F. Zuleger); Verimag, Grenoble, France (Dr. R. Iosif); IRIF, Paris, France (Prof. A. Bouajjani, Dr. M. Sighireanu).

   Within this topic, the student may also actively participate in various grant projects conducted within the VeriFIT group.

   Supervisor: Rogalewicz Adam, doc. Mgr., Ph.D.

4. Automated Techniques for Requirements-Based Test Assessment

   Evaluation of the completeness of test suites with respect to software requirements represents a fundamental challenge in the verification of critical systems, particularly in regulated domains. Current practice is largely based on manual assessment, which is time-consuming and error-prone. This dissertation will focus on the research of formal and automated methods for analyzing requirement coverage by test cases, with an emphasis on the semantic aspects of requirements and their implicit behavior.

   The goal of the dissertation is to propose a formal framework for representing requirements and test cases, and to define objective coverage criteria derived from the semantics of requirements (e.g., combinations of conditions, temporal constraints, sequences of actions, or state transitions). On this basis, methods for computing the coverage level of a test suite and for identifying uncovered semantic cases will be investigated.

   The research will focus in particular on the following areas:

   • design and systematic comparison of formal models suitable for representing different classes of requirements and testing artifacts, including analysis of their expressive power and analyzability;
   • formalization of the notion of requirement coverage and derivation of corresponding coverage obligations from the formal semantics of requirements;
   • design of algorithms for analyzing test suite coverage over formal representations, including identification of uncovered cases and their linkage to specific parts of the requirements;
   • investigation of methods for transforming informal requirements and test cases into formal models, including the use of natural language processing techniques and large language models, with emphasis on the reliability and limitations of these approaches;
   • experimental evaluation of the proposed methods on realistic datasets, including assessment of accuracy, completeness, and scalability.

   The dissertation will also include the design and implementation of prototype tools supporting selected aspects of the proposed approach, along with their validation on examples from embedded and avionics systems. Emphasis will be placed on integrating formal methods, software testing, and automated analysis techniques, with the aim of contributing to systematic and repeatable evaluation of test suite quality.

   Supervisor: Rogalewicz Adam, doc. Mgr., Ph.D.

5. Cognitive Load Mitigation in Human-Robot Collaboration using Mixed Reality

   The objective of this thesis is to investigate the cognitive load and stress levels experienced by workers during direct collaboration with robotic systems. The research utilises Mixed Reality (MR) technologies and Digital Twins to perform an in-depth analysis of factors affecting the human psyche in automated environments. The primary task is to identify the causes of stress across various industrial scenarios and propose methods to prevent or mitigate these negative effects through an adaptive interface.

   The work bridges empirical testing on real robotic workstations with their digital twins, allowing for the safe simulation and optimisation of critical situations. To quantify workload, biometric signals (e.g., heart rate, eye-tracking, or brain activity) will be used to provide feedback for dynamically adjusting the MR environment. By leveraging augmented reality for navigation and diminished reality to filter out environmental distractions, the research aims to create a workspace that adapts in real-time to the operator's current needs and cognitive capacity.

   The contribution of this research lies in defining standards for ergonomic and safe human-robot interaction in the spirit of Industry 5.0. The results will provide the industry with procedures for deploying advanced technologies that increase production efficiency while also protecting employees' mental health and long-term well-being. This approach is essential for the future integration of complex robotics into standard workflows without risking cognitive overload.

   Supervisor: Beran Vítězslav, doc. Ing., Ph.D.

6. Computer-Aided Creativity

   The aim of this dissertation is to conduct research in the field of generative artificial intelligence—whether it involves diffusion and adversarial models for video generation, generative text models for story creation, the automatic generation of computer code and music, the representation of knowledge in physics and chemistry, the promotion of scientific creativity, or a combination of all these approaches. The work will focus on addressing issues related to human interaction with generated intermediate results, the natural labeling of individual parts and concepts so that it is possible to build upon ongoing results, and the development of methods for modifying datasets and learning procedures to address societal issues associated with the creative models being developed—questions of model fairness, bias, and the incorporation of concepts of so-called responsible artificial intelligence.

   Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

7. Detection and Identification of Cryptocurrency Users from Network Traffic

   Motivation:

   Cryptocurrencies are generally used for the nearly instantaneous (i.e., until recorded on the blockchain), decentralized (i.e., without any regulatory authority), and (pseudo)anonymous transfer of cryptoassets (whether coins or tokens) between participants. These and many other characteristics stemming from the inherent use of cryptography are also the reason why cryptocurrencies are an integral part of cybercriminal activities (e.g., extortion via ransomware, purchases/sales on dark markets, investment scams).

   The majority of cryptocurrencies operate as a peer-to-peer network of nodes that mutually update and maintain a replica of the blockchain. Nodes communicate with each other using standardized protocols (e.g., the Bitcoin protocol [1]). Based on messages sent via the protocol, one can infer the state of a node (e.g., the creation of a new transaction) and also collect related identifiers (e.g., IP address, cryptocurrency address, or transaction hash). Furthermore, there are other overlay networks built on top of existing cryptocurrency networks (e.g., the Lightning Network [2]), where the same principles can be applied to interconnect information from them (similar to miner detection [3]), with the ultimate goal being the deanonymization of cryptocurrency users and their addresses. This dissertation could thus make a significant contribution to law enforcement agencies in identifying users of cryptocurrency addresses and linking their identities in the blockchain and network worlds.

   Objective:

   The aim of this work is to develop and validate new methods for analyzing cryptocurrency-related network traffic in order to deanonymize users and their associated identifiers.

   Expected phases of the project:

   1. Research on the use of cryptocurrencies in cybercrime in collaboration with law enforcement agencies.
   2. Analysis of existing approaches to the detection and analysis of cryptocurrency traffic and the use of this data to identify cryptocurrency users.
   3. Implementation of new approaches, deployment, and subsequent verification of their functionality in practice.
   4. Evaluation of the results achieved.

   Project:

   The work will thematically build on the results of earlier projects such as TARZAN and MIXER. The research efforts and expected outputs will be part of the COW (https://www.fit.vut.cz/research/project/c36822/) and COP IDEA (https://www.fit.vut.cz/research/project/c38120/) projects. Funding is primarily expected from the COW project.

   References:

   [1] P2P Network. In: Bitcoin Developer Guide [online]. Bitcoin.org [accessed 2026-04-07]. Available from: https://developer.bitcoin.org/devguide/p2p_network.html

   [2] POON, J. and DRYJA, T. The Bitcoin Lightning Network: Scalable Off-Chain Instant Payments [online]. Draft v0.5.9.2. 2016-01-14 [accessed 2026-04-07]. Available from: https://lightning.network/lightning-network-paper.pdf

   [3] VESELÝ, V.; ŽÁDNÍK, M. How to detect cryptocurrency miners? By traffic forensics!. Digital Investigation, 2019, vol. 31, no. 31, p. 1-14. ISSN: 1742-2876.

   Supervisor: Veselý Vladimír, Ing., Ph.D.

8. Detection and Prevention of Deepfake Attacks on Images of Human Faces

   The student will study techniques used in deepfake attacks on images and videos of human faces. The doctoral study is linked to the TrustedFace project (in collaboration with Innovatrics). The goal is to investigate existing deepfake attacks on images of human faces aimed at impersonating identities, quantify their severity, and propose possible defenses – both for detecting attacks and preventing them. The student will need to gain in-depth familiarity with available datasets of human faces, techniques for face identification and verification in images and video, and deepfake attack techniques. The ultimate goal of the study and doctoral work is to develop defenses against deepfake attacks and to formulate recommendations for systems that utilize face-based identification and verification.

   Supervisor: Herout Adam, prof. Ing., Ph.D.

9. Efficient finite automata for automated reasoning

   We will develop techniques for working with finite automata with applications in automatic reasoning, verification, and decision procedures for logics. Although automata are a fundamental concept of computer science with many applications and strong theoretical foundations, their practical use remains an active and challenging research area.

   The field is strongly motivated by applications in verification and testing, regular pattern search, artificial intelligence and inference, SMT-solving, system design, and automated synthesis. For example, the most efficient algorithms for regular pattern matching are automata-based, yet extending them to features such as backreferences, repetition counters, or look-arounds remains open. String solving has applications in program verification, particularly in security analysis of web applications (e.g., SQL injection, XSS), cloud access policies, and models of critical systems. Another challenge is efficient decision procedures for integer arithmetic and related logics. Regular model checking enables verification of infinite-state systems such as programs with dynamic data structures or communication protocols. Automated techniques can also support program synthesis (e.g., interfaces or drivers).

   These areas raise fundamental theoretical questions about decidability and complexity. For instance, how to model backreferences using automata? What classes of string-manipulating or data-structure programs are decidable, and at what cost?

   From a practical perspective, key challenges include efficient implementation of automata algorithms and mitigating state explosion. This involves compact representations via minimization, abstraction, approximation, and extensions such as alternating, symbolic, counter, or register automata. Their theoretical properties and efficient implementation for real-world problems remain open research directions.

   The supervision team includes Doc. Holík (VeriFIT), Dr. Lengál, Doc. Rogalewicz, Doc. Češka, and Prof. Vojnar, doctor Smrčka. The group is internationally recognized, with extensive publications, software tools, awards, and collaborations (e.g., Microsoft Research, Academia Sinica, Uppsala University). We also collaborate with industry (Red Hat, Honeywell). PhD graduates have strong prospects in both academia and industry, and students are supported through successful grant funding.

   Supervisor: Holík Lukáš, doc. Mgr., Ph.D.

10. Efficient HW accelerators for DNN inference

    Artificial convolutional and transformer neural networks (NNs) have recently been widely used in computationally intensive tasks such as classification, prediction, and recognition. The aim of this dissertation will be to propose methods for optimizing NN inference on embedded devices. The task will be to reduce the energy consumption of the output using advanced techniques, such as finding suitable data and weights representations, network architectures, optimizing computation scheduling on existing platforms, or implementing parts of the NN computation in-house. This research falls within the scope of topics addressed by the EvoAI Hardware research group.

    Supervisor: Mrázek Vojtěch, doc. Ing., Ph.D.

11. Efficient Technology for Dealing with Logics and Automata (Not Only) for Formal Analysis and Verification

    Different types of logics and automata are among the most fundamental objects studied and applied in computer science for decades. Nevertheless, there are many unsatisfactorily solved problems in this area, and new, exciting problems related to ever new applications of logics and automata are constantly emerging (e.g., in the formal verification of finite and infinite state systems with various advanced control or data structures, in decision procedures, in program or hardware synthesis, in quantum program verification, or even in methods for efficient search in various types of data or network traffic).

    The subject of the dissertation will primarily be the development of the state of the art in efficient work with various logics (e.g. over pointer structures, strings, various arithmetic, temporal logics, etc.). To this end, approaches based on different types of decision diagrams, automata, but also e.g. approaches based on the existence of a model of bounded size or on efficient reductions between different types of logical theories will be investigated. In connection with this, methods for efficient work with decision diagrams and different types of automata (automata over words, trees, infinite words, automata with counters, registers, etc.) will be developed. The work will include theoretical research as well as prototype implementation of the proposed techniques and their experimental evaluation.

    The work will be solved in collaboration with the VeriFIT team working on the development of techniques for working with logics and automata and their applications at FIT BUT. There is also the possibility of close cooperation with various foreign partners of VeriFIT: Academia Sinica, Taiwan (Prof. Y. Chen); TU Vienna, Austria (Assoc. F. Zuleger); LSV, ENS Paris-Saclay (Assoc. M. Sighireanu); IRIF, Paris, France (Assoc. A. Bouajjani, Assoc. P. Habermehl, Assoc. C. Enea), Verimag, Grenoble, France (Assoc. R. Iosif); Uppsala University, Sweden (prof. P.A. Abdulla, prof. B. Jonsson); or RPTU, Kaiserslautern, Germany (prof. A.W. Lin).

    Supervisor: Lengál Ondřej, doc. Ing., Ph.D.

12. Embedded Systems for Video/Signal Processing

    The topic focuses embedded image, video and/or signal processing. Its main goal is to research capabilities of "smart" and "small" units that have such features that allow for their applications requiring smyll, hidden, distributed, low power, mechanically or climatically stressed systems suitable of processing of some signal input. Exploitation of such systems is perspective and wide and also client/server and/or cloud systems. The units themselves can be based on CPU/DSP/GPU/AI accelerator, programmable hardware, or their combination. Smart cameras can be considered as well. Applications of interest include:

    • classification of images or objects using machine learning (AI) using traditional methods or through deep convolution networks neural network or similar approaches (e.g. for industrial quality inspection, etc.),
    • parallel analysis of signal(s) and video (e.g. for robust detection of occurrence of object in industrial or surveillance applications),
    • modern algorithms of video, image, and/or signal exploiting "client/server" or "cloud" (with focus on the technlogy) suitable e.g. for mobile technology and/or embedded systems,
    • other similar topics can be individually consulted and considered.

    A possibility exists in collaboration on grant projects, especially the newly submitted TAČR, HE, ECSEL ones (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

13. Energy-aware Embedded Inteligence

    The aim of this dissertation is to investigate models of embedded intelligence that explicitly account for the energy consumption of specific operations and optimise their performance based on specific constraints imposed by individual devices or the entire system. It will also include implementing selected models on suitable hardware for international projects in which the supervisor is involved.

    Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

14. Evaluating Vulnerabilities and Capabilities of Large Language Models

    Despite the impressive performance of large language models (LLMs), these models are still not well understood. A lot of effort is dedicated to evaluating the capabilities of LLMs and creating guardrails to prevent potentially harmful behaviours. However, small prompt variations still significantly alter the behaviour of LLMs, they often fail in tasks that are similar to the ones seen in training data but contain a small change, or they can be easily used to generate confident misinformation. As the large language models increasingly mediate access to information and are increasingly used by expert and non-expert users, it is essential to develop rigorous methods for evaluating their true capabilities and identifying the sources of their vulnerabilities.

    The goal of this research is to build a deeper and more reliable understanding of the LLM behaviour. The first possible research direction involves designing a benchmark that can accurately measure both advanced capabilities (such as reasoning, planning, understanding, math or multilingual capabilities) and common failure modes and vulnerabilities (including hallucination, brittleness, tendency or ease of generating problematic content such as misinformation). The second possible research direction leverages mechanistic interpretability to study the internal structures that drive the LLM behaviours. These tools will be used to investigate why specific vulnerabilities arise, how the capabilities arise, and whether targeted interventions can reduce failures while maintaining core capabilities.

    By integrating robust behavioural benchmarking with mechanistic insights, the project aims to produce a more principled understanding of how LLMs work and how they can be made safer and more reliable. The application domain includes (but are not limited to) low-resource tasks, multilingual understanding, misinformation and other problematic behaviour detection (narrative detection) or even LLM user simulation.

    Relevant publications:

    • Vykopal, I., Pikuliak, M., Srba, I., Moro, R., Macko, D., and Bielikova, M., 2023. 2024. Disinformation Capabilities of Large Language Models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14830–14847, Bangkok, Thailand. Association for Computational Linguistics. https://aclanthology.org/2024.acl-long.793/
    • Zugecova, A., Macko, D., Srba, I., Moro, R., Kopal, J., Marcinčinová, K., and Mesarčík, M., 2025. Evaluation of LLM Vulnerabilities to Being Misused for Personalized Disinformation Generation. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 780–797, Vienna, Austria. Association for Computational Linguistics. https://aclanthology.org/2025.acl-long.38/

    The research will be performed at the Kempelen Institute of Intelligent Technologies (KInIT, https://kinit.sk) in Bratislava in cooperation with researchers from highly respected research units. A combined (external) form of study and full employment at KInIT is expected.

    Supervisor: Bieliková Mária, prof. Ing., Ph.D.

15. Evolutionary machine learning for automated digital circuit design

    .

    Supervisor: Sekanina Lukáš, prof. Ing., Ph.D.

16. Explainable Artificial Intelligence

    Explainable AI (XAI) centres on enhancing trust, interpretability, and usability of complex models. Key areas include creating human-centred explanations, establishing objective evaluation metrics for, and balancing accuracy with interpretability. Recently, the research has shifted toward interactive explanations, evaluating how AI models handle data and bias, transparently communicating result uncertainty, and evaluating the quality of the generated explanation.

    The research will focus on human-centred and cognitive aspects, aligning AI explanations with human psychology and cognitive abilities, highlighting differences between human-human and AI-human explanations, developing interactive, dialogue-based explanations that allow users to ask "what-if" questions and explore model behaviours, and objectively measuring the quality, understandability, and effectiveness of an explanation for a user. The initial application domain will correspond to fact-checking in the context of the fight against desinformation.

    Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

17. Formal Methods in Design and Analysis of Quantum Programs

    Development in the field of quantum computers is moving forward unstoppably. However, for the effective use of quantum hardware, it is necessary to have the right support at the software level, i.e., quantum algorithms for solving problems, as well as tools for compiling, optimizing and synthesizing quantum programs, mapping them to a specific quantum technology, mapping them to a fault tolerant protocol, analyzing, simulating, and debugging quantum programs, etc.

    The subject of the dissertation will primarily be the development of current knowledge in the field of formal methods in quantum program design. The main focus here will be on the use of automata theory, automatic reasoning, and decision diagrams. In connection with this, methods for effective work with decision diagrams and various types of automata (automata over words, trees, infinite words, automata with counters, registers, etc.) will also be developed. The work will include both theoretical research and prototype implementation of the proposed techniques and their experimental evaluation.

    The work will be carried out in collaboration with the VeriFIT team at FIT BUT, which develops techniques for working with logics and automata and their applications. There is also the possibility of close cooperation with various foreign VeriFIT partners: Academia Sinica, Taiwan (prof. Y. Chen); Uppsala University, Sweden (prof. P.A. Abdulla, prof. R. S. Thinniyam); or Leiden Institute for Advanced Computer Science, Leiden, the Netherlands (prof. A. Laarman).

    Supervisor: Lengál Ondřej, doc. Ing., Ph.D.

18. Genetic programming and LLM for automated circuit and program synthesis

    .

    Supervisor: Sekanina Lukáš, prof. Ing., Ph.D.

19. Hardware architectures and methods for mapping neural networks onto FPGAs

    In recent years, neural networks (NNs) have achieved breakthrough results in tasks such as image recognition, natural language processing, object detection, and predictive control. Growing interest in deploying NNs in applications with strict response requirements—such as autonomous vehicles, industrial robotics, stock trading, or network traffic monitoring—is creating pressure to reduce inference latency to the range of hundreds of nanoseconds to microseconds.

    While general-purpose computing platforms (CPUs, GPUs) offer high flexibility, they face physical limitations imposed by data transfer between memory and the processor, as well as high power consumption. In contrast, FPGAs (Field-Programmable Gate Arrays) allow data paths, parallelism, and computational accuracy to be fully tailored to a specific network architecture. Thanks to their reconfigurability, deterministic latency, and low power consumption, FPGAs represent a promising platform for NN acceleration, particularly in edge computing and embedded systems.

    Nevertheless, key research questions remain open: how to efficiently map various network topologies onto a limited chip area, how to minimize latency while maintaining acceptable accuracy, and how to automate and generalize this process for different families of NNs and FPGAs.

    Objectives of the Dissertation

    1. Analysis of the Current State – A systematic review of existing approaches to mapping NNs onto FPGAs, identifying their limitations in terms of latency, throughput, and energy efficiency.
    2. Design of new hardware architectures – Research and design of specialized processing elements and memory hierarchies optimized for low-latency inference of convolutional (CNN), recurrent (RNN), and transformer architectures.
    3. FPGA-Targeted Network Quantization and Pruning Methods – Research into bit-width reduction techniques (mixed-precision quantization), structural pruning, and model compression tailored to the specifics of FPGA architecture, with the goal of minimizing computational and memory requirements while maintaining accuracy.
    4. Automated mapping (HLS and others) – Design and verification of methodologies or tools for the automatic generation of FPGA implementations from high-level NN descriptions, with latency optimization as the primary criterion (extensions of HLS approaches, e.g., Vitis HLS, or direct synthesis at the RTL level).
    5. Experimental validation – Implementation and measurement of the proposed architectures on real FPGA hardware (e.g., Xilinx/AMD Ultrascale+, Intel/Altera Agilex) and comparison with reference solutions (GPUs, existing FPGA frameworks such as hls4ml, finn).
    6. Generalization and transferability – Formulation of general principles and recommendations for the design of low-latency FPGA NN accelerators applicable across various domains.

    Expected benefits

    • New insights into the relationship between NN architecture, mapping strategy, and achieved inference latency on FPGAs.
    • Design and verification of hardware architectures that surpass the current state of the art in terms of latency.
    • Publication of results in peer-reviewed international journals and at conferences (FPL, FCCM, DAC, DATE, ICCAD).

    Supervisor: Martínek Tomáš, doc. Ing., Ph.D.

20. High-Performance Traffic Shaping for Multi-Gigabit Edge Networks

    The transition of residential and access network infrastructure to multi-gigabit plans for end customers has invalidated many of the assumptions on which the network traffic quality of service subsystem within the Linux kernel is based. Traditional algorithms, such as Hierarchical Token Bucket, suffer from a significant performance degradation at speeds of several gigabits due to the need to use global locks and cache locality violations.

    This dissertation focuses on the optimization and design of new algorithms and architectures to meet network service level agreement (SLA) parameters, even in a hierarchical network topology. The thesis aims to explore new mechanisms and lock-free algorithms with distributed state management at the individual processor level. The research will examine trade-offs between flexibility and efficiency in the Linux kernel so that the proposed algorithms can meet the required parameters even at speeds of several gigabits per second while ensuring the necessary level of quality of service.

    Supervisor: Matoušek Petr, doc. Ing., Ph.D., M.A.

21. Homographic Navigation – augmented reality for capturing precise images of objects

    The aim of this work will be to develop the technique of "homographic navigation," a new concept being developed under the guidance of the doctoral program supervisor. The PhD student will develop computer vision algorithms and related machine learning techniques. Model training for homographic navigation must be minimally supervised. Both the learning algorithms and the inference and use in the application must be fast and resource-efficient. The project also involves prototyping and developing an augmented reality application that will be used for data collection, user testing, and other related tasks. Homographic navigation technology is intended to have immediate industrial applications, so the doctoral study will also include the development of all essential techniques through to the production phase.

    Supervisor: Herout Adam, prof. Ing., Ph.D.

22. Human in the loop (HIL) in the adaptation of AI systems

    This PhD topic is driven by the needs of users for which the privacy of data is crucial and concentrates on improving AI (especially speech recognition) system in case the user is able to identify and correct systems errors. The topic includes proper evaluations of HIL systems, selection of data to be proposed for correction and actual fine-tuning / adaptation techniques working with large pre-trained models. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

23. Making Machine Learning Alive

    Towards "Living" Machine

    Bio-Inspired Recurrent Architectures: Integrating Top-Down Feedback  for Adaptive Inference in Non-Stationary Environments.

    9Problem Statement

    Unlike biological systems, which continue to learn and adapt during the inference process, standard ML models cannot self-correct or reconfigure their internal processing streams when faced with novel or shifting data distributions. This lack of dynamic inference leads to catastrophic failure in changing environments.

    Research Objectives

    The goal of this research is to move beyond "dead" feedforward architectures by implementing three key biological principles:

    • Reciprocal Recurrence: Replace unidirectional chains with parallel processing streams and massive top-down feedback loops.
    • Active Inference Learning: Develop mechanisms that allow the model to evaluate the quality of its own extracted information in real-time.
    • Gating: Implement weighting of lower-level neuron outputs, controlled by higher-level feedback to filter irrelevant data and focus on contextually salient features.

    Expected Contribution

    This research aims to deliver a blueprint for a new class of Self-Correcting Artificial Intelligence. By mimicking the "dynamic and alive" nature of biological perceptual systems, we can create AI that doesn't just "predict," but actively "perceives" and adjusts its strategies in real-time—essentially turning inference from a static snapshot into a continuous, adaptive process.

    Supervisor: Heřmanský Hynek, prof. Ing., Dr. Eng.

24. Methods for knowledge extraction and representation from heterogeneous sources to improve the performance and interpretability of AI systems

    Currently, there is a dynamic development of artificial intelligence methods, particularly large language models, which enable advanced processing and interpretation of extensive unstructured data. However, efficient knowledge mining from heterogeneous data sources and their transformation into a form suitable for training and improving AI systems remains an open problem. The quality and structure of input knowledge fundamentally influence the resulting accuracy, robustness, and interpretability of these systems.

    With the increasing complexity of data and requirements for automated decision-making, there is a growing need to develop methods that allow for systematic extraction, representation, and utilization of knowledge across various domains. This issue has both significant theoretical contribution and broad application potential in the field of intelligent agents and decision support systems.
    The aim of the doctoral thesis is to design, implement, and experimentally verify new approaches to knowledge mining, representation, and utilization from heterogeneous data sources to improve the performance of AI systems based on language models.

    1. Analysis of the current state of knowledge
       • Systematic review of knowledge mining methods and their application in AI
       • Identification of the limits of existing approaches when working with unstructured and multimodal data
    2. Design of knowledge mining methods
       • Development of new approaches for extracting relevant knowledge from diverse data sources (text, tables, documents, logs)
       • Design of methods for integrating structured and unstructured data
    3. Representation and integration of knowledge into AI systems
       • Design of suitable knowledge representations (e.g., knowledge graphs, embedding representations)
       • Optimization of knowledge utilization in language models (e.g., retrieval-augmented generation, fine-tuning)
    4. Experimental verification of the proposed approaches
       • Implementation of the proposed methods
       • Testing on real-world and publicly available datasets
       • Comparison with existing methods using appropriate metrics
    5. Verification of transferability and generalizability
       • Application of the proposed methods to various domains and use cases
       • Evaluation of the robustness and adaptability of the approaches
    6. Publication and application outputs
       • Publication of results in scientific journals and at conferences
       • Creation of prototype tools or frameworks

    At the same time, the thesis will be conceived generally so that the results are transferable to other areas where working with knowledge and data is key, such as enterprise information systems, decision support systems, autonomous AI agents, or applications in industry, healthcare, or public administration.

    Supervisor: Burget Radek, doc. Ing., Ph.D.

25. Modeling Latent Representations of Speech Attributes Using Generative Models

    This work focuses on the study and design of generative neural models for learning latent representations of speech signals, with an emphasis on the effective capture and disentanglement of individual speech attributes such as speaker identity, emotion, prosody, and articulation. The goal is to develop models that enable the disentanglement of these attributes in the latent space and their subsequent controllable manipulation during speech synthesis. The research will include an analysis of existing approaches (e.g., variational autoencoders, generative adversarial networks, and diffusion models), the design of novel methods, and their experimental evaluation on real-world speech data. The expected outcome is an improvement in the quality and flexibility of generated speech signals, as well as a deeper understanding of speech representation in generative models.

    Supervisor: Burget Lukáš, doc. Ing., Ph.D.

26. Modern Algorithms of Computer Graphics

    The topic concerns algorithms of computer graphics and image synthesis. Its main goal is to research new algorithms so that their features and application possibilities are better understood so that they are improved or newly created. If suitable, it is possible to work on various platforms, includeing parallel CPUs, such as x86/64, ARM, Xeon PHI, GPU, etc. or other cores in CUDA, OpenCl, VHDL, etc. Algorithms of interest include:

    • rendering using selected computer graphics methods (such as ray tracing, photon mapping, direct rendering of "point clouds", etc.),
    • modeling of scenes and redering using artificial intelligence, including image synthesis using neural netowrks (especially CNNs),
    • processing and rendering of "lightfield" images, their acquisition, or possibly compression, reconstruction of 3D scenes from images and/or video, eventually also fusing with other sensors, such as LIDAR or RADAR,
    • modern algorithms of geometry suitable for applications in cpmputer graphics and perhaps also 3D printing,
    • emerging algorithms of 3D synthesis, holography, wavelet transform, frequency transform, etc.

    After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

    Collaboration on grant projects, such as TACR, HE, ECSEL (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

27. Multimodal Analysis of Forensic Evidence Using Artificial Intelligence Methods

    This dissertation will focus on the development of advanced methods for the analysis of forensic evidence using multimodal data, particularly image and textual information. The rapid advancement of artificial intelligence methods—especially deep neural networks and large pre-trained models—significantly expands the possibilities of automated processing and interpretation of forensic materials. This progress contributes to increased efficiency and objectivity in forensic examination, including applications within the Customs Administration and other security agencies. A key challenge lies in the robust analysis of unconventional and limited datasets.
    The research will address multimodal analysis of forensic evidence with an emphasis on practical applicability, specifically:
    (1) the design of new methods for analyzing unconventional image data, such as X-ray images from inspection systems used by the Customs Administration;
    (2) the development of methods for data augmentation and synthetic dataset expansion to improve the robustness and generalization of models in detecting risk-related objects; and
    (3) the integration of computer vision methods with natural language processing techniques for joint evaluation of visual evidence and related textual documentation.
    The overall aim of the dissertation is to propose and validate a robust multimodal approach to forensic evidence analysis applicable in practice within the Customs Administration and other security agencies.

    Supervisor: Hanáček Petr, doc. Dr. Ing.

28. Network configuration generated using LLM models

    Large LLM language models are now applied in various areas where they perform common and repetitive activities. LLM support can also be used in network design and configuration of network devices, where they can generate a configuration in the domain language of a given manufacturer (e.g. Cisco IOS, Juniper OS, etc.) based on instructions in the recommended language. However, configuration generation requires accuracy and the ability to detect possible errors that arise from inaccuracies in input, hallucinations of LLM models, etc.

    The assigned topic develops the principle called Intent-based Networking, where we will use LLM language models and a system of cooperating agents to find a procedure for implementing the desired network behavior and generate the corresponding configurations for specific network devices. In order to be able to deploy the given procedure in real-world networks, it is necessary to design a method for verifying the generated configuration by combining static and dynamic verification, which will verify whether the given configuration meets the required properties.  

    Supervisor: Matoušek Petr, doc. Ing., Ph.D., M.A.

29. Neurosymbolic methods for safe agent control

    The subject of the dissertation is the development of theoretical foundations and algorithms for safe agent control using neurosymbolic AI, which combines deep reinforcement learning methods and symbolic formal methods. Emphasis will be placed on approaches that guarantee safe and explainable solutions. The dissertation will focus on the synthesis of decision trees and finite-state controllers, as well as on monitoring methods and runtime verification for agent systems. The dissertation will also examine model-learning methods and model abstractions enabling the utilization of formal methods. The work will focus on the use of these methods in the field of agent control in environments with probabilistic dynamics and partial observability, and on a thorough evaluation of the applicability of these methods. The results of this work will contribute to the field of autonomous systems.

    Research on neurosymbolic control methods is currently receiving considerable attention in the fields of artificial intelligence and formal methods, as evidenced by the focus of a number of top conferences (e.g. AAAI, AAMAS, CAV, or TACAS).

    The work will be carried out in cooperation with the VeriFIT team under the leadership of doc. M. Češka. Furthermore, close cooperation is expected with the group of Prof. N. Jansen (ERC Starting grant investigator), Bochum University, Germany, Prof. D. Parker, Oxford University, UK and Prof. S. Junges Radboud University Nijmegen, Netherlands.

    In case of a responsible approach and high-quality results, there is a possibility of participation in grant projects (e.g. Czech Science Foundation or Horizon Europe projects).

    Supervisor: Češka Milan, doc. RNDr., Ph.D.

30. New approaches to the numerical solution of partial differential equations

    Many technical problems are described by the partial differential equations. These equations are very difficult to solve analytically and many approaches exist that allow for the numerical approximation of the solution. The aim of this PhD thesis is to study the existing approaches to the numerical solution of the partial differential equations and develop new approaches to their solution. The student will mostly concentrate on the solution of the system in time domain, becoming acquainted with the wide range of methods used for the numerical solution of the differential equations.

    Supervisor: Jaroš Jiří, prof. Ing., Ph.D.

31. Non-Traditional Visal Computing Algorithms

    The topic of the work includes algorithms of comptuer graphics, image, and video, i.e. "Visual Computing", such as HDR (High Dynamic Range) image, multispectral image, stereoimage, possibly also image augmented with material features, temperature, etc. The goal is to understand the features and possibilites but also applications, to analyze the algorithms in-depth, prepare new and improve them. The possible algorithms include:
    
    

    • Acquisition of HDR image and video, its compression, assembling from several standard images, videos, etc.,
    • acquisition, processing, and visualization of mutispectral images (images with more than three spectral components),
    • processing of stereo images and other images containing the depth information using both structural light illumination or no structural light,
    • algorithms suitable for mobile technology and/or embedded systems focusing the above types of iamges, or images acquired from drones,
    • metods of visualization of HDR images and video, tone-mapping, real-time tone mapping,
    • algorithms and applications exploiting Wavelet, frequency, and/or other similar transformations.

     

    After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

    Collaboration on grant projects, such as TACR, MPO, HE, ECSEL (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

32. Performance of voice deepfake detection by humans and machines

    The work will start with getting familiar with the basics of the problem of voice deep fake detection (DFD), terminology, available techniques, data and challenges (AVSpoof, WildSpoof), with the history and state of the art techniques and tools for speaker recognition (wespeaker toolkit), with state of the art techniques and tools for personalized text to speech (pTTS) synthesis and voice conversion. The topic will then concentrate on the evaluation of human and machine performance in DFD, while concentrating on the motivation aspect of human subjects to simulate true attacks on people. The PhD will then progress in both (1) technical (advances in deepfake generation and detection technology) and (2) human-aspects. The work counts on collaboration with specialists from psychology and sociology. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

33. Performance of voice deepfake detection by humans and machines

    The work will start with getting familiar with the basics of the problem of voice deep fake detection (DFD), terminology, available techniques, data and challenges (AVSpoof, WildSpoof), with the history and state of the art techniques and tools for speaker recognition (wespeaker toolkit), with state of the art techniques and tools for personalized text to speech (pTTS) synthesis and voice conversion. The topic will then concentrate on the evaluation of human and machine performance in DFD, while concentrating on the motivation aspect of human subjects to simulate true attacks on people. The PhD will then progress in both (1) technical (advances in deepfake generation and detection technology) and (2) human-aspects. The work counts on collaboration with specialists from psychology and sociology. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

34. Self-evaluation Machine Learning

    Research Gap: The "Blind" Inference Problem

    Current AI suffers from "silent failure." When a model encounters data outside its training distribution, it still provides an output—often with high overconfidence—because it lacks a biological Self-Evaluation Module. It cannot "feel" its own confusion. Biological systems survive because they evaluate the quality of extracted information during the act of perceiving, allowing them to switch strategies instantly.

    Research Focus: The SEM

    This research proposes the development of an independent Self-Evaluation Module (SEM) that sits atop standard processing layers. Its primary functions include:

    • Confidence Synthesis: Moving beyond simple "Softmax" probability to multi-factor reliability metrics (e.g., entropy analysis, reconstruction error, or latent space stability).
    • The "Surprise" Trigger: Detecting when new data contradicts internal world models (novelty detection) to trigger an "adaptation phase" during a single forward pass.
    • Efferent Control Gating: Using the evaluation score to send top-down signals that "gate" (turn on/off) specific lower-level neurons, effectively reconfiguring the model's focus in real-time.

    Key Research Questions

    • How can a model calculate the "quality" of its own inference without access to ground-truth labels?
    • How do we prevent "feedback loops" where a self-evaluating model becomes stuck in a recursive state of doubt?

    Impact

    By perfecting the Self-Evaluation Module, we move toward AI that knows when it doesn't know. This is critical for safety-critical applications like autonomous driving or medical diagnostics, where a machine’s ability to flag its own uncertainty is as important as the prediction itself.

    Supervisor: Heřmanský Hynek, prof. Ing., Dr. Eng.

35. Speech recognition and data processing for ATC-pilot communication in aviation

    This PhD topic is focused on speech recognition and data processing for ATC-pilot communication in aviation. It will cover all components of automatic speech recognition (ASR), i.e. data processing, acoustic model, vocabulary (including special aviation terminology), language model as well as interaction with sources of meta-data (radar information, airport). Special attention will be paid to the use of large pre-trained model, code-switching, reliable language identification from short segments and adaptation to the conditions of given airport with crawled data.

    Supervisor: Černocký Jan, prof. Dr. Ing.

36. String constraint solving in applications

    We will develop techniques for string constraint solving, i.e., automated reasoning about logics over strings, with applications in verification, testing, and security. String solving has recently reached a level of maturity that enables precise analysis of string-manipulating programs in real-world systems .

    The student will contribute to advancing the core technology, focusing on improving scalability and expressiveness of string solvers, in particular within the SMT-solving framework. This includes automata-based techniques, integration with other theories (e.g., integers, arrays), and handling advanced string operations such as replacements, concatenation, and string-number conversions.

    A key part of the work will be applying string solving in challenging domains such as security analysis of web applications, analysis of access control policies, and testing of safety-critical systems. These applications require reasoning about complex relations between string values that cannot be handled by traditional testing or static analysis techniques.

    The student will also participate in a joint project with Embraer, focusing on applications of SMT solving, string solving, and related techniques in industrial settings. This includes advanced testing, static analysis, and verification of real-world systems, and may involve a stronger engineering component, including implementation and integration into existing toolchains.

    The research will be conducted within the VeriFIT group (L. Holík, O. Lengál, A. Rogalewicz, M. Češka, T. Vojnar), an internationally recognized team with strong publication record, industrial collaborations, and experience in developing state-of-the-art verification tools. PhD students benefit from close collaboration with leading academic and industrial partners and strong career prospects in both academia and industry.

    Supervisor: Holík Lukáš, doc. Mgr., Ph.D.

37. The impact of the life cycle of large language models on cybersecurity

    In recent years, there has been an increase in the use of neural networks for generating synthetic content, which goes hand in hand with the rise of new cybersecurity challenges. Generative models can have various impacts on cybersecurity, ranging from positive to negative.A significant area is the security of deployment and operation of generative models, primarily large language models.

    The goal of this thesis is to identify problem areas in a selected field of LLM deployment and operation (e.g., model inference attacks, model theft, or information theft) and analyze new trends, approaches, defenses, and their characteristics, impacts, and potential applications. The work should then propose new protection methods based on the analysis and research on the state of security for the selected areas.

    Recommended areas of work focus:
    Attacks and defenses in the side channel domain of large language models
    Attacks and defenses in the inference domain of large language models
    Defences in the area of adversarial attacks on large language models

    Participation in relevant international conferences and publication in peer-reviewed or scientific journals are expected.

    Supervisor: Malinka Kamil, doc. Mgr., Ph.D.

38. Unconventional Wave Propagation Models

    The aim of the dissertation is to develop unconventional models for acoustic wave propagation. The work will focus on advanced acoustic propagators, Gaussian beam models, and decompositions of spectral methods. The goal is to create a hybrid model that uses k-Wave for low frequencies, while employing advanced propagators for high frequencies. This will significantly reduce both spatial and temporal complexity. In addition to designing the algorithm itself, we will also work on the efficient implementation of these models on clusters of GPU workstations.

    Supervisor: Jaroš Jiří, prof. Ing., Ph.D.

39. Visual Geo-Localization and Augmented Reality

    The project deals with geo-localization of images and videos in unknown environments using computer vision and computer graphics methods. The aim is to investigate and develop new image registration techniques (with geo-localized image database or 3D terrain model). The goal is an efficient implementation of proposed methods on mobile devices as well as search for additional applications in the area of image processing, computational photography, and augmented reality.

    • Further information: http://cadik.posvete.cz/locate/
    • Contact: http://cadik.posvete.cz/
    • Cooperation and research visits with leading research labs are possible (Adobe Research, USA, MPII Saarbrücken, Germany, Disney Research Zurich, Switzerland, INRIA Bordeaux, France)

    Supervisor: Čadík Martin, prof. Ing., Ph.D.

2. round (applications submitted from 01.06.2026 to 30.08.2026)

1. Advanced algorithms of Video, Image, and/or Signal processing

   The topic concerns algorithms of image, video, and/or signal processing. Its main goal is to research and in-depth analyze existing algorithms and discover new ones so that they have desirable features and so that they are possible to efficiently implement. Such efficient implementation can be but does not necessarily have to be part of the work but it is important to prepare the algorithms so that they can be efficiently implemented e.g. in CPU, in GPU, in AI accelratorrs, embeded systems, even in combination with FPGA, in extremely low power systems, or in other environments. It is possible to exploit algorithms of artificial intelligence, such as neural networks, especially CNNs, LLM/VLM. The application possibilities of the algorithms are also important and the application can be but does not have to be part of the work. The algorithms/applications of interest include:

   • recognition of scene contents, events, and general semantics of video sequences (such as identification of traffic situations, identification in scenes in moview, action identification, etc.),
   • classification of video sequences using machine learning (AI) through deep convolution networks neural network or similar approaches (e.g. for industrial quality inspection, object of scene characteristics search, etc.), possibly in combination with object tracking in video using modern methods,
   • parallel analysis of video and signal (e.g. for detection of coincidence of occurrence of object in video and characteristic signal shape in surveillance applications), fusion of video and sognals,
   • modern algorithms of video, image, and/or signal exploiting "client/server" or "cloud" approaches suitable e.g. for mobile technology and/or embedded systems,
   • algorithms of video compression and analysis through frequency or wavelet transforms or similar methods...

   After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

   Collaboration on grant projects, such as TACR, MPO, H2020, ECSEL (employment expected).

   Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

2. Agentic AI in Chip Design and Other Parts of Electronic Design Automation

   Research in Agentic AI for chip design and Electronic Design Automation (EDA) is rapidly shifting from passive, single-point tool optimisation to autonomous, goal-driven workflows. Agentic AI systems use reasoning agents to plan, orchestrate EDA tools, learn from outcomes, and collaborate to manage multi-step tasks throughout the semiconductor design lifecycle. 

   The goal of the research is to bring AI-native autonomy to various stages of the frontend and backend components of the IC design, including RTL, verification and physical design. An initial step will be to adapt and use specialised models trained on Verilog, netlists, and layout data to understand the semantic structure of electronic circuits, translate requirements into synthesizable code, and iterate automatically using compiler feedback in the context of the RISC-V Automotive Processor explored in the Chips JU project Rigoletto.

   Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

3. Analysis of programs with complex data structure

   Software projects that employ dynamic data structures in combination with pointers are highly prone to memory-related errors (such as null pointer dereferencing, memory leaks, etc.). At the same time, such code is often system-level code used in operating system kernels, shared libraries, or interpreters of higher-level programming languages.

   The aim of this dissertation is to build upon existing techniques for the analysis of programs with dynamic memory, in particular those based on Separation Logic and bi-abduction. The work may focus on the C/C++ programming language or another suitable language. It will build on prior work by members of the VeriFIT research group and by Dr. F. Zuleger from TU Wien.

   An alternative direction of the dissertation is to build upon current techniques for automated complexity analysis, with the goal of analyzing open code and/or code with complex data structures. This direction will also build on the work of the VeriFIT group, as well as Dr. F. Zuleger from TU Wien and Dr. M. Sinn from the University of Applied Sciences St. Pölten.

   The dissertation will be carried out in collaboration with the VeriFIT team at the Faculty of Information Technology, Brno University of Technology, which focuses on program verification techniques for programs with complex data structures (in particular Dr. L. Holík, Prof. T. Vojnar, Ing. T. Dacík, Ing. V. Šoková). With a responsible approach and high-quality results, there is an opportunity to participate in research grant projects, including international ones. There is also the possibility of close collaboration with various international partners of VeriFIT: TU Wien, Austria (Dr. F. Zuleger); Verimag, Grenoble, France (Dr. R. Iosif); IRIF, Paris, France (Prof. A. Bouajjani, Dr. M. Sighireanu).

   Within this topic, the student may also actively participate in various grant projects conducted within the VeriFIT group.

   Supervisor: Rogalewicz Adam, doc. Mgr., Ph.D.

4. Automated Techniques for Requirements-Based Test Assessment

   Evaluation of the completeness of test suites with respect to software requirements represents a fundamental challenge in the verification of critical systems, particularly in regulated domains. Current practice is largely based on manual assessment, which is time-consuming and error-prone. This dissertation will focus on the research of formal and automated methods for analyzing requirement coverage by test cases, with an emphasis on the semantic aspects of requirements and their implicit behavior.

   The goal of the dissertation is to propose a formal framework for representing requirements and test cases, and to define objective coverage criteria derived from the semantics of requirements (e.g., combinations of conditions, temporal constraints, sequences of actions, or state transitions). On this basis, methods for computing the coverage level of a test suite and for identifying uncovered semantic cases will be investigated.

   The research will focus in particular on the following areas:

   • design and systematic comparison of formal models suitable for representing different classes of requirements and testing artifacts, including analysis of their expressive power and analyzability;
   • formalization of the notion of requirement coverage and derivation of corresponding coverage obligations from the formal semantics of requirements;
   • design of algorithms for analyzing test suite coverage over formal representations, including identification of uncovered cases and their linkage to specific parts of the requirements;
   • investigation of methods for transforming informal requirements and test cases into formal models, including the use of natural language processing techniques and large language models, with emphasis on the reliability and limitations of these approaches;
   • experimental evaluation of the proposed methods on realistic datasets, including assessment of accuracy, completeness, and scalability.

   The dissertation will also include the design and implementation of prototype tools supporting selected aspects of the proposed approach, along with their validation on examples from embedded and avionics systems. Emphasis will be placed on integrating formal methods, software testing, and automated analysis techniques, with the aim of contributing to systematic and repeatable evaluation of test suite quality.

   Supervisor: Rogalewicz Adam, doc. Mgr., Ph.D.

5. Cognitive Load Mitigation in Human-Robot Collaboration using Mixed Reality

   The objective of this thesis is to investigate the cognitive load and stress levels experienced by workers during direct collaboration with robotic systems. The research utilises Mixed Reality (MR) technologies and Digital Twins to perform an in-depth analysis of factors affecting the human psyche in automated environments. The primary task is to identify the causes of stress across various industrial scenarios and propose methods to prevent or mitigate these negative effects through an adaptive interface.

   The work bridges empirical testing on real robotic workstations with their digital twins, allowing for the safe simulation and optimisation of critical situations. To quantify workload, biometric signals (e.g., heart rate, eye-tracking, or brain activity) will be used to provide feedback for dynamically adjusting the MR environment. By leveraging augmented reality for navigation and diminished reality to filter out environmental distractions, the research aims to create a workspace that adapts in real-time to the operator's current needs and cognitive capacity.

   The contribution of this research lies in defining standards for ergonomic and safe human-robot interaction in the spirit of Industry 5.0. The results will provide the industry with procedures for deploying advanced technologies that increase production efficiency while also protecting employees' mental health and long-term well-being. This approach is essential for the future integration of complex robotics into standard workflows without risking cognitive overload.

   Supervisor: Beran Vítězslav, doc. Ing., Ph.D.

6. Computer-Aided Creativity

   The aim of this dissertation is to conduct research in the field of generative artificial intelligence—whether it involves diffusion and adversarial models for video generation, generative text models for story creation, the automatic generation of computer code and music, the representation of knowledge in physics and chemistry, the promotion of scientific creativity, or a combination of all these approaches. The work will focus on addressing issues related to human interaction with generated intermediate results, the natural labeling of individual parts and concepts so that it is possible to build upon ongoing results, and the development of methods for modifying datasets and learning procedures to address societal issues associated with the creative models being developed—questions of model fairness, bias, and the incorporation of concepts of so-called responsible artificial intelligence.

   Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

7. Detection and Identification of Cryptocurrency Users from Network Traffic

   Motivation:

   Cryptocurrencies are generally used for the nearly instantaneous (i.e., until recorded on the blockchain), decentralized (i.e., without any regulatory authority), and (pseudo)anonymous transfer of cryptoassets (whether coins or tokens) between participants. These and many other characteristics stemming from the inherent use of cryptography are also the reason why cryptocurrencies are an integral part of cybercriminal activities (e.g., extortion via ransomware, purchases/sales on dark markets, investment scams).

   The majority of cryptocurrencies operate as a peer-to-peer network of nodes that mutually update and maintain a replica of the blockchain. Nodes communicate with each other using standardized protocols (e.g., the Bitcoin protocol [1]). Based on messages sent via the protocol, one can infer the state of a node (e.g., the creation of a new transaction) and also collect related identifiers (e.g., IP address, cryptocurrency address, or transaction hash). Furthermore, there are other overlay networks built on top of existing cryptocurrency networks (e.g., the Lightning Network [2]), where the same principles can be applied to interconnect information from them (similar to miner detection [3]), with the ultimate goal being the deanonymization of cryptocurrency users and their addresses. This dissertation could thus make a significant contribution to law enforcement agencies in identifying users of cryptocurrency addresses and linking their identities in the blockchain and network worlds.

   Objective:

   The aim of this work is to develop and validate new methods for analyzing cryptocurrency-related network traffic in order to deanonymize users and their associated identifiers.

   Expected phases of the project:

   1. Research on the use of cryptocurrencies in cybercrime in collaboration with law enforcement agencies.
   2. Analysis of existing approaches to the detection and analysis of cryptocurrency traffic and the use of this data to identify cryptocurrency users.
   3. Implementation of new approaches, deployment, and subsequent verification of their functionality in practice.
   4. Evaluation of the results achieved.

   Project:

   The work will thematically build on the results of earlier projects such as TARZAN and MIXER. The research efforts and expected outputs will be part of the COW (https://www.fit.vut.cz/research/project/c36822/) and COP IDEA (https://www.fit.vut.cz/research/project/c38120/) projects. Funding is primarily expected from the COW project.

   References:

   [1] P2P Network. In: Bitcoin Developer Guide [online]. Bitcoin.org [accessed 2026-04-07]. Available from: https://developer.bitcoin.org/devguide/p2p_network.html

   [2] POON, J. and DRYJA, T. The Bitcoin Lightning Network: Scalable Off-Chain Instant Payments [online]. Draft v0.5.9.2. 2016-01-14 [accessed 2026-04-07]. Available from: https://lightning.network/lightning-network-paper.pdf

   [3] VESELÝ, V.; ŽÁDNÍK, M. How to detect cryptocurrency miners? By traffic forensics!. Digital Investigation, 2019, vol. 31, no. 31, p. 1-14. ISSN: 1742-2876.

   Supervisor: Veselý Vladimír, Ing., Ph.D.

8. Detection and Prevention of Deepfake Attacks on Images of Human Faces

   The student will study techniques used in deepfake attacks on images and videos of human faces. The doctoral study is linked to the TrustedFace project (in collaboration with Innovatrics). The goal is to investigate existing deepfake attacks on images of human faces aimed at impersonating identities, quantify their severity, and propose possible defenses – both for detecting attacks and preventing them. The student will need to gain in-depth familiarity with available datasets of human faces, techniques for face identification and verification in images and video, and deepfake attack techniques. The ultimate goal of the study and doctoral work is to develop defenses against deepfake attacks and to formulate recommendations for systems that utilize face-based identification and verification.

   Supervisor: Herout Adam, prof. Ing., Ph.D.

9. Efficient finite automata for automated reasoning

   We will develop techniques for working with finite automata with applications in automatic reasoning, verification, and decision procedures for logics. Although automata are a fundamental concept of computer science with many applications and strong theoretical foundations, their practical use remains an active and challenging research area.

   The field is strongly motivated by applications in verification and testing, regular pattern search, artificial intelligence and inference, SMT-solving, system design, and automated synthesis. For example, the most efficient algorithms for regular pattern matching are automata-based, yet extending them to features such as backreferences, repetition counters, or look-arounds remains open. String solving has applications in program verification, particularly in security analysis of web applications (e.g., SQL injection, XSS), cloud access policies, and models of critical systems. Another challenge is efficient decision procedures for integer arithmetic and related logics. Regular model checking enables verification of infinite-state systems such as programs with dynamic data structures or communication protocols. Automated techniques can also support program synthesis (e.g., interfaces or drivers).

   These areas raise fundamental theoretical questions about decidability and complexity. For instance, how to model backreferences using automata? What classes of string-manipulating or data-structure programs are decidable, and at what cost?

   From a practical perspective, key challenges include efficient implementation of automata algorithms and mitigating state explosion. This involves compact representations via minimization, abstraction, approximation, and extensions such as alternating, symbolic, counter, or register automata. Their theoretical properties and efficient implementation for real-world problems remain open research directions.

   The supervision team includes Doc. Holík (VeriFIT), Dr. Lengál, Doc. Rogalewicz, Doc. Češka, and Prof. Vojnar, doctor Smrčka. The group is internationally recognized, with extensive publications, software tools, awards, and collaborations (e.g., Microsoft Research, Academia Sinica, Uppsala University). We also collaborate with industry (Red Hat, Honeywell). PhD graduates have strong prospects in both academia and industry, and students are supported through successful grant funding.

   Supervisor: Holík Lukáš, doc. Mgr., Ph.D.

10. Efficient HW accelerators for DNN inference

    Artificial convolutional and transformer neural networks (NNs) have recently been widely used in computationally intensive tasks such as classification, prediction, and recognition. The aim of this dissertation will be to propose methods for optimizing NN inference on embedded devices. The task will be to reduce the energy consumption of the output using advanced techniques, such as finding suitable data and weights representations, network architectures, optimizing computation scheduling on existing platforms, or implementing parts of the NN computation in-house. This research falls within the scope of topics addressed by the EvoAI Hardware research group.

    Supervisor: Mrázek Vojtěch, doc. Ing., Ph.D.

11. Efficient Technology for Dealing with Logics and Automata (Not Only) for Formal Analysis and Verification

    Different types of logics and automata are among the most fundamental objects studied and applied in computer science for decades. Nevertheless, there are many unsatisfactorily solved problems in this area, and new, exciting problems related to ever new applications of logics and automata are constantly emerging (e.g., in the formal verification of finite and infinite state systems with various advanced control or data structures, in decision procedures, in program or hardware synthesis, in quantum program verification, or even in methods for efficient search in various types of data or network traffic).

    The subject of the dissertation will primarily be the development of the state of the art in efficient work with various logics (e.g. over pointer structures, strings, various arithmetic, temporal logics, etc.). To this end, approaches based on different types of decision diagrams, automata, but also e.g. approaches based on the existence of a model of bounded size or on efficient reductions between different types of logical theories will be investigated. In connection with this, methods for efficient work with decision diagrams and different types of automata (automata over words, trees, infinite words, automata with counters, registers, etc.) will be developed. The work will include theoretical research as well as prototype implementation of the proposed techniques and their experimental evaluation.

    The work will be solved in collaboration with the VeriFIT team working on the development of techniques for working with logics and automata and their applications at FIT BUT. There is also the possibility of close cooperation with various foreign partners of VeriFIT: Academia Sinica, Taiwan (Prof. Y. Chen); TU Vienna, Austria (Assoc. F. Zuleger); LSV, ENS Paris-Saclay (Assoc. M. Sighireanu); IRIF, Paris, France (Assoc. A. Bouajjani, Assoc. P. Habermehl, Assoc. C. Enea), Verimag, Grenoble, France (Assoc. R. Iosif); Uppsala University, Sweden (prof. P.A. Abdulla, prof. B. Jonsson); or RPTU, Kaiserslautern, Germany (prof. A.W. Lin).

    Supervisor: Lengál Ondřej, doc. Ing., Ph.D.

12. Embedded Systems for Video/Signal Processing

    The topic focuses embedded image, video and/or signal processing. Its main goal is to research capabilities of "smart" and "small" units that have such features that allow for their applications requiring smyll, hidden, distributed, low power, mechanically or climatically stressed systems suitable of processing of some signal input. Exploitation of such systems is perspective and wide and also client/server and/or cloud systems. The units themselves can be based on CPU/DSP/GPU/AI accelerator, programmable hardware, or their combination. Smart cameras can be considered as well. Applications of interest include:

    • classification of images or objects using machine learning (AI) using traditional methods or through deep convolution networks neural network or similar approaches (e.g. for industrial quality inspection, etc.),
    • parallel analysis of signal(s) and video (e.g. for robust detection of occurrence of object in industrial or surveillance applications),
    • modern algorithms of video, image, and/or signal exploiting "client/server" or "cloud" (with focus on the technlogy) suitable e.g. for mobile technology and/or embedded systems,
    • other similar topics can be individually consulted and considered.

    A possibility exists in collaboration on grant projects, especially the newly submitted TAČR, HE, ECSEL ones (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

13. Energy-aware Embedded Inteligence

    The aim of this dissertation is to investigate models of embedded intelligence that explicitly account for the energy consumption of specific operations and optimise their performance based on specific constraints imposed by individual devices or the entire system. It will also include implementing selected models on suitable hardware for international projects in which the supervisor is involved.

    Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

14. Evaluating Vulnerabilities and Capabilities of Large Language Models

    Despite the impressive performance of large language models (LLMs), these models are still not well understood. A lot of effort is dedicated to evaluating the capabilities of LLMs and creating guardrails to prevent potentially harmful behaviours. However, small prompt variations still significantly alter the behaviour of LLMs, they often fail in tasks that are similar to the ones seen in training data but contain a small change, or they can be easily used to generate confident misinformation. As the large language models increasingly mediate access to information and are increasingly used by expert and non-expert users, it is essential to develop rigorous methods for evaluating their true capabilities and identifying the sources of their vulnerabilities.

    The goal of this research is to build a deeper and more reliable understanding of the LLM behaviour. The first possible research direction involves designing a benchmark that can accurately measure both advanced capabilities (such as reasoning, planning, understanding, math or multilingual capabilities) and common failure modes and vulnerabilities (including hallucination, brittleness, tendency or ease of generating problematic content such as misinformation). The second possible research direction leverages mechanistic interpretability to study the internal structures that drive the LLM behaviours. These tools will be used to investigate why specific vulnerabilities arise, how the capabilities arise, and whether targeted interventions can reduce failures while maintaining core capabilities.

    By integrating robust behavioural benchmarking with mechanistic insights, the project aims to produce a more principled understanding of how LLMs work and how they can be made safer and more reliable. The application domain includes (but are not limited to) low-resource tasks, multilingual understanding, misinformation and other problematic behaviour detection (narrative detection) or even LLM user simulation.

    Relevant publications:

    • Vykopal, I., Pikuliak, M., Srba, I., Moro, R., Macko, D., and Bielikova, M., 2023. 2024. Disinformation Capabilities of Large Language Models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14830–14847, Bangkok, Thailand. Association for Computational Linguistics. https://aclanthology.org/2024.acl-long.793/
    • Zugecova, A., Macko, D., Srba, I., Moro, R., Kopal, J., Marcinčinová, K., and Mesarčík, M., 2025. Evaluation of LLM Vulnerabilities to Being Misused for Personalized Disinformation Generation. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 780–797, Vienna, Austria. Association for Computational Linguistics. https://aclanthology.org/2025.acl-long.38/

    The research will be performed at the Kempelen Institute of Intelligent Technologies (KInIT, https://kinit.sk) in Bratislava in cooperation with researchers from highly respected research units. A combined (external) form of study and full employment at KInIT is expected.

    Supervisor: Bieliková Mária, prof. Ing., Ph.D.

15. Evolutionary machine learning for automated digital circuit design

    .

    Supervisor: Sekanina Lukáš, prof. Ing., Ph.D.

16. Explainable Artificial Intelligence

    Explainable AI (XAI) centres on enhancing trust, interpretability, and usability of complex models. Key areas include creating human-centred explanations, establishing objective evaluation metrics for, and balancing accuracy with interpretability. Recently, the research has shifted toward interactive explanations, evaluating how AI models handle data and bias, transparently communicating result uncertainty, and evaluating the quality of the generated explanation.

    The research will focus on human-centred and cognitive aspects, aligning AI explanations with human psychology and cognitive abilities, highlighting differences between human-human and AI-human explanations, developing interactive, dialogue-based explanations that allow users to ask "what-if" questions and explore model behaviours, and objectively measuring the quality, understandability, and effectiveness of an explanation for a user. The initial application domain will correspond to fact-checking in the context of the fight against desinformation.

    Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

17. Formal Methods in Design and Analysis of Quantum Programs

    Development in the field of quantum computers is moving forward unstoppably. However, for the effective use of quantum hardware, it is necessary to have the right support at the software level, i.e., quantum algorithms for solving problems, as well as tools for compiling, optimizing and synthesizing quantum programs, mapping them to a specific quantum technology, mapping them to a fault tolerant protocol, analyzing, simulating, and debugging quantum programs, etc.

    The subject of the dissertation will primarily be the development of current knowledge in the field of formal methods in quantum program design. The main focus here will be on the use of automata theory, automatic reasoning, and decision diagrams. In connection with this, methods for effective work with decision diagrams and various types of automata (automata over words, trees, infinite words, automata with counters, registers, etc.) will also be developed. The work will include both theoretical research and prototype implementation of the proposed techniques and their experimental evaluation.

    The work will be carried out in collaboration with the VeriFIT team at FIT BUT, which develops techniques for working with logics and automata and their applications. There is also the possibility of close cooperation with various foreign VeriFIT partners: Academia Sinica, Taiwan (prof. Y. Chen); Uppsala University, Sweden (prof. P.A. Abdulla, prof. R. S. Thinniyam); or Leiden Institute for Advanced Computer Science, Leiden, the Netherlands (prof. A. Laarman).

    Supervisor: Lengál Ondřej, doc. Ing., Ph.D.

18. Genetic programming and LLM for automated circuit and program synthesis

    .

    Supervisor: Sekanina Lukáš, prof. Ing., Ph.D.

19. Hardware architectures and methods for mapping neural networks onto FPGAs

    In recent years, neural networks (NNs) have achieved breakthrough results in tasks such as image recognition, natural language processing, object detection, and predictive control. Growing interest in deploying NNs in applications with strict response requirements—such as autonomous vehicles, industrial robotics, stock trading, or network traffic monitoring—is creating pressure to reduce inference latency to the range of hundreds of nanoseconds to microseconds.

    While general-purpose computing platforms (CPUs, GPUs) offer high flexibility, they face physical limitations imposed by data transfer between memory and the processor, as well as high power consumption. In contrast, FPGAs (Field-Programmable Gate Arrays) allow data paths, parallelism, and computational accuracy to be fully tailored to a specific network architecture. Thanks to their reconfigurability, deterministic latency, and low power consumption, FPGAs represent a promising platform for NN acceleration, particularly in edge computing and embedded systems.

    Nevertheless, key research questions remain open: how to efficiently map various network topologies onto a limited chip area, how to minimize latency while maintaining acceptable accuracy, and how to automate and generalize this process for different families of NNs and FPGAs.

    Objectives of the Dissertation

    1. Analysis of the Current State – A systematic review of existing approaches to mapping NNs onto FPGAs, identifying their limitations in terms of latency, throughput, and energy efficiency.
    2. Design of new hardware architectures – Research and design of specialized processing elements and memory hierarchies optimized for low-latency inference of convolutional (CNN), recurrent (RNN), and transformer architectures.
    3. FPGA-Targeted Network Quantization and Pruning Methods – Research into bit-width reduction techniques (mixed-precision quantization), structural pruning, and model compression tailored to the specifics of FPGA architecture, with the goal of minimizing computational and memory requirements while maintaining accuracy.
    4. Automated mapping (HLS and others) – Design and verification of methodologies or tools for the automatic generation of FPGA implementations from high-level NN descriptions, with latency optimization as the primary criterion (extensions of HLS approaches, e.g., Vitis HLS, or direct synthesis at the RTL level).
    5. Experimental validation – Implementation and measurement of the proposed architectures on real FPGA hardware (e.g., Xilinx/AMD Ultrascale+, Intel/Altera Agilex) and comparison with reference solutions (GPUs, existing FPGA frameworks such as hls4ml, finn).
    6. Generalization and transferability – Formulation of general principles and recommendations for the design of low-latency FPGA NN accelerators applicable across various domains.

    Expected benefits

    • New insights into the relationship between NN architecture, mapping strategy, and achieved inference latency on FPGAs.
    • Design and verification of hardware architectures that surpass the current state of the art in terms of latency.
    • Publication of results in peer-reviewed international journals and at conferences (FPL, FCCM, DAC, DATE, ICCAD).

    Supervisor: Martínek Tomáš, doc. Ing., Ph.D.

20. High-Performance Traffic Shaping for Multi-Gigabit Edge Networks

    The transition of residential and access network infrastructure to multi-gigabit plans for end customers has invalidated many of the assumptions on which the network traffic quality of service subsystem within the Linux kernel is based. Traditional algorithms, such as Hierarchical Token Bucket, suffer from a significant performance degradation at speeds of several gigabits due to the need to use global locks and cache locality violations.

    This dissertation focuses on the optimization and design of new algorithms and architectures to meet network service level agreement (SLA) parameters, even in a hierarchical network topology. The thesis aims to explore new mechanisms and lock-free algorithms with distributed state management at the individual processor level. The research will examine trade-offs between flexibility and efficiency in the Linux kernel so that the proposed algorithms can meet the required parameters even at speeds of several gigabits per second while ensuring the necessary level of quality of service.

    Supervisor: Matoušek Petr, doc. Ing., Ph.D., M.A.

21. Homographic Navigation – augmented reality for capturing precise images of objects

    The aim of this work will be to develop the technique of "homographic navigation," a new concept being developed under the guidance of the doctoral program supervisor. The PhD student will develop computer vision algorithms and related machine learning techniques. Model training for homographic navigation must be minimally supervised. Both the learning algorithms and the inference and use in the application must be fast and resource-efficient. The project also involves prototyping and developing an augmented reality application that will be used for data collection, user testing, and other related tasks. Homographic navigation technology is intended to have immediate industrial applications, so the doctoral study will also include the development of all essential techniques through to the production phase.

    Supervisor: Herout Adam, prof. Ing., Ph.D.

22. Human in the loop (HIL) in the adaptation of AI systems

    This PhD topic is driven by the needs of users for which the privacy of data is crucial and concentrates on improving AI (especially speech recognition) system in case the user is able to identify and correct systems errors. The topic includes proper evaluations of HIL systems, selection of data to be proposed for correction and actual fine-tuning / adaptation techniques working with large pre-trained models. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

23. Making Machine Learning Alive

    Towards "Living" Machine

    Bio-Inspired Recurrent Architectures: Integrating Top-Down Feedback  for Adaptive Inference in Non-Stationary Environments.

    9Problem Statement

    Unlike biological systems, which continue to learn and adapt during the inference process, standard ML models cannot self-correct or reconfigure their internal processing streams when faced with novel or shifting data distributions. This lack of dynamic inference leads to catastrophic failure in changing environments.

    Research Objectives

    The goal of this research is to move beyond "dead" feedforward architectures by implementing three key biological principles:

    • Reciprocal Recurrence: Replace unidirectional chains with parallel processing streams and massive top-down feedback loops.
    • Active Inference Learning: Develop mechanisms that allow the model to evaluate the quality of its own extracted information in real-time.
    • Gating: Implement weighting of lower-level neuron outputs, controlled by higher-level feedback to filter irrelevant data and focus on contextually salient features.

    Expected Contribution

    This research aims to deliver a blueprint for a new class of Self-Correcting Artificial Intelligence. By mimicking the "dynamic and alive" nature of biological perceptual systems, we can create AI that doesn't just "predict," but actively "perceives" and adjusts its strategies in real-time—essentially turning inference from a static snapshot into a continuous, adaptive process.

    Supervisor: Heřmanský Hynek, prof. Ing., Dr. Eng.

24. Methods for knowledge extraction and representation from heterogeneous sources to improve the performance and interpretability of AI systems

    Currently, there is a dynamic development of artificial intelligence methods, particularly large language models, which enable advanced processing and interpretation of extensive unstructured data. However, efficient knowledge mining from heterogeneous data sources and their transformation into a form suitable for training and improving AI systems remains an open problem. The quality and structure of input knowledge fundamentally influence the resulting accuracy, robustness, and interpretability of these systems.

    With the increasing complexity of data and requirements for automated decision-making, there is a growing need to develop methods that allow for systematic extraction, representation, and utilization of knowledge across various domains. This issue has both significant theoretical contribution and broad application potential in the field of intelligent agents and decision support systems.
    The aim of the doctoral thesis is to design, implement, and experimentally verify new approaches to knowledge mining, representation, and utilization from heterogeneous data sources to improve the performance of AI systems based on language models.

    1. Analysis of the current state of knowledge
       • Systematic review of knowledge mining methods and their application in AI
       • Identification of the limits of existing approaches when working with unstructured and multimodal data
    2. Design of knowledge mining methods
       • Development of new approaches for extracting relevant knowledge from diverse data sources (text, tables, documents, logs)
       • Design of methods for integrating structured and unstructured data
    3. Representation and integration of knowledge into AI systems
       • Design of suitable knowledge representations (e.g., knowledge graphs, embedding representations)
       • Optimization of knowledge utilization in language models (e.g., retrieval-augmented generation, fine-tuning)
    4. Experimental verification of the proposed approaches
       • Implementation of the proposed methods
       • Testing on real-world and publicly available datasets
       • Comparison with existing methods using appropriate metrics
    5. Verification of transferability and generalizability
       • Application of the proposed methods to various domains and use cases
       • Evaluation of the robustness and adaptability of the approaches
    6. Publication and application outputs
       • Publication of results in scientific journals and at conferences
       • Creation of prototype tools or frameworks

    At the same time, the thesis will be conceived generally so that the results are transferable to other areas where working with knowledge and data is key, such as enterprise information systems, decision support systems, autonomous AI agents, or applications in industry, healthcare, or public administration.

    Supervisor: Burget Radek, doc. Ing., Ph.D.

25. Modeling Latent Representations of Speech Attributes Using Generative Models

    This work focuses on the study and design of generative neural models for learning latent representations of speech signals, with an emphasis on the effective capture and disentanglement of individual speech attributes such as speaker identity, emotion, prosody, and articulation. The goal is to develop models that enable the disentanglement of these attributes in the latent space and their subsequent controllable manipulation during speech synthesis. The research will include an analysis of existing approaches (e.g., variational autoencoders, generative adversarial networks, and diffusion models), the design of novel methods, and their experimental evaluation on real-world speech data. The expected outcome is an improvement in the quality and flexibility of generated speech signals, as well as a deeper understanding of speech representation in generative models.

    Supervisor: Burget Lukáš, doc. Ing., Ph.D.

26. Modern Algorithms of Computer Graphics

    The topic concerns algorithms of computer graphics and image synthesis. Its main goal is to research new algorithms so that their features and application possibilities are better understood so that they are improved or newly created. If suitable, it is possible to work on various platforms, includeing parallel CPUs, such as x86/64, ARM, Xeon PHI, GPU, etc. or other cores in CUDA, OpenCl, VHDL, etc. Algorithms of interest include:

    • rendering using selected computer graphics methods (such as ray tracing, photon mapping, direct rendering of "point clouds", etc.),
    • modeling of scenes and redering using artificial intelligence, including image synthesis using neural netowrks (especially CNNs),
    • processing and rendering of "lightfield" images, their acquisition, or possibly compression, reconstruction of 3D scenes from images and/or video, eventually also fusing with other sensors, such as LIDAR or RADAR,
    • modern algorithms of geometry suitable for applications in cpmputer graphics and perhaps also 3D printing,
    • emerging algorithms of 3D synthesis, holography, wavelet transform, frequency transform, etc.

    After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

    Collaboration on grant projects, such as TACR, HE, ECSEL (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

27. Multimodal Analysis of Forensic Evidence Using Artificial Intelligence Methods

    This dissertation will focus on the development of advanced methods for the analysis of forensic evidence using multimodal data, particularly image and textual information. The rapid advancement of artificial intelligence methods—especially deep neural networks and large pre-trained models—significantly expands the possibilities of automated processing and interpretation of forensic materials. This progress contributes to increased efficiency and objectivity in forensic examination, including applications within the Customs Administration and other security agencies. A key challenge lies in the robust analysis of unconventional and limited datasets.
    The research will address multimodal analysis of forensic evidence with an emphasis on practical applicability, specifically:
    (1) the design of new methods for analyzing unconventional image data, such as X-ray images from inspection systems used by the Customs Administration;
    (2) the development of methods for data augmentation and synthetic dataset expansion to improve the robustness and generalization of models in detecting risk-related objects; and
    (3) the integration of computer vision methods with natural language processing techniques for joint evaluation of visual evidence and related textual documentation.
    The overall aim of the dissertation is to propose and validate a robust multimodal approach to forensic evidence analysis applicable in practice within the Customs Administration and other security agencies.

    Supervisor: Hanáček Petr, doc. Dr. Ing.

28. Network configuration generated using LLM models

    Large LLM language models are now applied in various areas where they perform common and repetitive activities. LLM support can also be used in network design and configuration of network devices, where they can generate a configuration in the domain language of a given manufacturer (e.g. Cisco IOS, Juniper OS, etc.) based on instructions in the recommended language. However, configuration generation requires accuracy and the ability to detect possible errors that arise from inaccuracies in input, hallucinations of LLM models, etc.

    The assigned topic develops the principle called Intent-based Networking, where we will use LLM language models and a system of cooperating agents to find a procedure for implementing the desired network behavior and generate the corresponding configurations for specific network devices. In order to be able to deploy the given procedure in real-world networks, it is necessary to design a method for verifying the generated configuration by combining static and dynamic verification, which will verify whether the given configuration meets the required properties.  

    Supervisor: Matoušek Petr, doc. Ing., Ph.D., M.A.

29. Neurosymbolic methods for safe agent control

    The subject of the dissertation is the development of theoretical foundations and algorithms for safe agent control using neurosymbolic AI, which combines deep reinforcement learning methods and symbolic formal methods. Emphasis will be placed on approaches that guarantee safe and explainable solutions. The dissertation will focus on the synthesis of decision trees and finite-state controllers, as well as on monitoring methods and runtime verification for agent systems. The dissertation will also examine model-learning methods and model abstractions enabling the utilization of formal methods. The work will focus on the use of these methods in the field of agent control in environments with probabilistic dynamics and partial observability, and on a thorough evaluation of the applicability of these methods. The results of this work will contribute to the field of autonomous systems.

    Research on neurosymbolic control methods is currently receiving considerable attention in the fields of artificial intelligence and formal methods, as evidenced by the focus of a number of top conferences (e.g. AAAI, AAMAS, CAV, or TACAS).

    The work will be carried out in cooperation with the VeriFIT team under the leadership of doc. M. Češka. Furthermore, close cooperation is expected with the group of Prof. N. Jansen (ERC Starting grant investigator), Bochum University, Germany, Prof. D. Parker, Oxford University, UK and Prof. S. Junges Radboud University Nijmegen, Netherlands.

    In case of a responsible approach and high-quality results, there is a possibility of participation in grant projects (e.g. Czech Science Foundation or Horizon Europe projects).

    Supervisor: Češka Milan, doc. RNDr., Ph.D.

30. New approaches to the numerical solution of partial differential equations

    Many technical problems are described by the partial differential equations. These equations are very difficult to solve analytically and many approaches exist that allow for the numerical approximation of the solution. The aim of this PhD thesis is to study the existing approaches to the numerical solution of the partial differential equations and develop new approaches to their solution. The student will mostly concentrate on the solution of the system in time domain, becoming acquainted with the wide range of methods used for the numerical solution of the differential equations.

    Supervisor: Jaroš Jiří, prof. Ing., Ph.D.

31. Non-Traditional Visal Computing Algorithms

    The topic of the work includes algorithms of comptuer graphics, image, and video, i.e. "Visual Computing", such as HDR (High Dynamic Range) image, multispectral image, stereoimage, possibly also image augmented with material features, temperature, etc. The goal is to understand the features and possibilites but also applications, to analyze the algorithms in-depth, prepare new and improve them. The possible algorithms include:
    
    

    • Acquisition of HDR image and video, its compression, assembling from several standard images, videos, etc.,
    • acquisition, processing, and visualization of mutispectral images (images with more than three spectral components),
    • processing of stereo images and other images containing the depth information using both structural light illumination or no structural light,
    • algorithms suitable for mobile technology and/or embedded systems focusing the above types of iamges, or images acquired from drones,
    • metods of visualization of HDR images and video, tone-mapping, real-time tone mapping,
    • algorithms and applications exploiting Wavelet, frequency, and/or other similar transformations.

     

    After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

    Collaboration on grant projects, such as TACR, MPO, HE, ECSEL (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

32. Performance of voice deepfake detection by humans and machines

    The work will start with getting familiar with the basics of the problem of voice deep fake detection (DFD), terminology, available techniques, data and challenges (AVSpoof, WildSpoof), with the history and state of the art techniques and tools for speaker recognition (wespeaker toolkit), with state of the art techniques and tools for personalized text to speech (pTTS) synthesis and voice conversion. The topic will then concentrate on the evaluation of human and machine performance in DFD, while concentrating on the motivation aspect of human subjects to simulate true attacks on people. The PhD will then progress in both (1) technical (advances in deepfake generation and detection technology) and (2) human-aspects. The work counts on collaboration with specialists from psychology and sociology. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

33. Performance of voice deepfake detection by humans and machines

    The work will start with getting familiar with the basics of the problem of voice deep fake detection (DFD), terminology, available techniques, data and challenges (AVSpoof, WildSpoof), with the history and state of the art techniques and tools for speaker recognition (wespeaker toolkit), with state of the art techniques and tools for personalized text to speech (pTTS) synthesis and voice conversion. The topic will then concentrate on the evaluation of human and machine performance in DFD, while concentrating on the motivation aspect of human subjects to simulate true attacks on people. The PhD will then progress in both (1) technical (advances in deepfake generation and detection technology) and (2) human-aspects. The work counts on collaboration with specialists from psychology and sociology. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

34. Self-evaluation Machine Learning

    Research Gap: The "Blind" Inference Problem

    Current AI suffers from "silent failure." When a model encounters data outside its training distribution, it still provides an output—often with high overconfidence—because it lacks a biological Self-Evaluation Module. It cannot "feel" its own confusion. Biological systems survive because they evaluate the quality of extracted information during the act of perceiving, allowing them to switch strategies instantly.

    Research Focus: The SEM

    This research proposes the development of an independent Self-Evaluation Module (SEM) that sits atop standard processing layers. Its primary functions include:

    • Confidence Synthesis: Moving beyond simple "Softmax" probability to multi-factor reliability metrics (e.g., entropy analysis, reconstruction error, or latent space stability).
    • The "Surprise" Trigger: Detecting when new data contradicts internal world models (novelty detection) to trigger an "adaptation phase" during a single forward pass.
    • Efferent Control Gating: Using the evaluation score to send top-down signals that "gate" (turn on/off) specific lower-level neurons, effectively reconfiguring the model's focus in real-time.

    Key Research Questions

    • How can a model calculate the "quality" of its own inference without access to ground-truth labels?
    • How do we prevent "feedback loops" where a self-evaluating model becomes stuck in a recursive state of doubt?

    Impact

    By perfecting the Self-Evaluation Module, we move toward AI that knows when it doesn't know. This is critical for safety-critical applications like autonomous driving or medical diagnostics, where a machine’s ability to flag its own uncertainty is as important as the prediction itself.

    Supervisor: Heřmanský Hynek, prof. Ing., Dr. Eng.

35. Speech recognition and data processing for ATC-pilot communication in aviation

    This PhD topic is focused on speech recognition and data processing for ATC-pilot communication in aviation. It will cover all components of automatic speech recognition (ASR), i.e. data processing, acoustic model, vocabulary (including special aviation terminology), language model as well as interaction with sources of meta-data (radar information, airport). Special attention will be paid to the use of large pre-trained model, code-switching, reliable language identification from short segments and adaptation to the conditions of given airport with crawled data.

    Supervisor: Černocký Jan, prof. Dr. Ing.

36. String constraint solving in applications

    We will develop techniques for string constraint solving, i.e., automated reasoning about logics over strings, with applications in verification, testing, and security. String solving has recently reached a level of maturity that enables precise analysis of string-manipulating programs in real-world systems .

    The student will contribute to advancing the core technology, focusing on improving scalability and expressiveness of string solvers, in particular within the SMT-solving framework. This includes automata-based techniques, integration with other theories (e.g., integers, arrays), and handling advanced string operations such as replacements, concatenation, and string-number conversions.

    A key part of the work will be applying string solving in challenging domains such as security analysis of web applications, analysis of access control policies, and testing of safety-critical systems. These applications require reasoning about complex relations between string values that cannot be handled by traditional testing or static analysis techniques.

    The student will also participate in a joint project with Embraer, focusing on applications of SMT solving, string solving, and related techniques in industrial settings. This includes advanced testing, static analysis, and verification of real-world systems, and may involve a stronger engineering component, including implementation and integration into existing toolchains.

    The research will be conducted within the VeriFIT group (L. Holík, O. Lengál, A. Rogalewicz, M. Češka, T. Vojnar), an internationally recognized team with strong publication record, industrial collaborations, and experience in developing state-of-the-art verification tools. PhD students benefit from close collaboration with leading academic and industrial partners and strong career prospects in both academia and industry.

    Supervisor: Holík Lukáš, doc. Mgr., Ph.D.

37. The impact of the life cycle of large language models on cybersecurity

    In recent years, there has been an increase in the use of neural networks for generating synthetic content, which goes hand in hand with the rise of new cybersecurity challenges. Generative models can have various impacts on cybersecurity, ranging from positive to negative.A significant area is the security of deployment and operation of generative models, primarily large language models.

    The goal of this thesis is to identify problem areas in a selected field of LLM deployment and operation (e.g., model inference attacks, model theft, or information theft) and analyze new trends, approaches, defenses, and their characteristics, impacts, and potential applications. The work should then propose new protection methods based on the analysis and research on the state of security for the selected areas.

    Recommended areas of work focus:
    Attacks and defenses in the side channel domain of large language models
    Attacks and defenses in the inference domain of large language models
    Defences in the area of adversarial attacks on large language models

    Participation in relevant international conferences and publication in peer-reviewed or scientific journals are expected.

    Supervisor: Malinka Kamil, doc. Mgr., Ph.D.

38. Unconventional Wave Propagation Models

    The aim of the dissertation is to develop unconventional models for acoustic wave propagation. The work will focus on advanced acoustic propagators, Gaussian beam models, and decompositions of spectral methods. The goal is to create a hybrid model that uses k-Wave for low frequencies, while employing advanced propagators for high frequencies. This will significantly reduce both spatial and temporal complexity. In addition to designing the algorithm itself, we will also work on the efficient implementation of these models on clusters of GPU workstations.

    Supervisor: Jaroš Jiří, prof. Ing., Ph.D.

39. Visual Geo-Localization and Augmented Reality

    The project deals with geo-localization of images and videos in unknown environments using computer vision and computer graphics methods. The aim is to investigate and develop new image registration techniques (with geo-localized image database or 3D terrain model). The goal is an efficient implementation of proposed methods on mobile devices as well as search for additional applications in the area of image processing, computational photography, and augmented reality.

    • Further information: http://cadik.posvete.cz/locate/
    • Contact: http://cadik.posvete.cz/
    • Cooperation and research visits with leading research labs are possible (Adobe Research, USA, MPII Saarbrücken, Germany, Disney Research Zurich, Switzerland, INRIA Bordeaux, France)

    Supervisor: Čadík Martin, prof. Ing., Ph.D.

1. round (applications submitted from 01.03.2026 to 31.05.2026)

1. Advanced algorithms of Video, Image, and/or Signal processing

   The topic concerns algorithms of image, video, and/or signal processing. Its main goal is to research and in-depth analyze existing algorithms and discover new ones so that they have desirable features and so that they are possible to efficiently implement. Such efficient implementation can be but does not necessarily have to be part of the work but it is important to prepare the algorithms so that they can be efficiently implemented e.g. in CPU, in GPU, in AI accelratorrs, embeded systems, even in combination with FPGA, in extremely low power systems, or in other environments. It is possible to exploit algorithms of artificial intelligence, such as neural networks, especially CNNs, LLM/VLM. The application possibilities of the algorithms are also important and the application can be but does not have to be part of the work. The algorithms/applications of interest include:

   • recognition of scene contents, events, and general semantics of video sequences (such as identification of traffic situations, identification in scenes in moview, action identification, etc.),
   • classification of video sequences using machine learning (AI) through deep convolution networks neural network or similar approaches (e.g. for industrial quality inspection, object of scene characteristics search, etc.), possibly in combination with object tracking in video using modern methods,
   • parallel analysis of video and signal (e.g. for detection of coincidence of occurrence of object in video and characteristic signal shape in surveillance applications), fusion of video and sognals,
   • modern algorithms of video, image, and/or signal exploiting "client/server" or "cloud" approaches suitable e.g. for mobile technology and/or embedded systems,
   • algorithms of video compression and analysis through frequency or wavelet transforms or similar methods...

   After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

   Collaboration on grant projects, such as TACR, MPO, H2020, ECSEL (employment expected).

   Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

2. Agentic AI in Chip Design and Other Parts of Electronic Design Automation

   Research in Agentic AI for chip design and Electronic Design Automation (EDA) is rapidly shifting from passive, single-point tool optimisation to autonomous, goal-driven workflows. Agentic AI systems use reasoning agents to plan, orchestrate EDA tools, learn from outcomes, and collaborate to manage multi-step tasks throughout the semiconductor design lifecycle. 

   The goal of the research is to bring AI-native autonomy to various stages of the frontend and backend components of the IC design, including RTL, verification and physical design. An initial step will be to adapt and use specialised models trained on Verilog, netlists, and layout data to understand the semantic structure of electronic circuits, translate requirements into synthesizable code, and iterate automatically using compiler feedback in the context of the RISC-V Automotive Processor explored in the Chips JU project Rigoletto.

   Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

3. Analysis of programs with complex data structure

   Software projects that employ dynamic data structures in combination with pointers are highly prone to memory-related errors (such as null pointer dereferencing, memory leaks, etc.). At the same time, such code is often system-level code used in operating system kernels, shared libraries, or interpreters of higher-level programming languages.

   The aim of this dissertation is to build upon existing techniques for the analysis of programs with dynamic memory, in particular those based on Separation Logic and bi-abduction. The work may focus on the C/C++ programming language or another suitable language. It will build on prior work by members of the VeriFIT research group and by Dr. F. Zuleger from TU Wien.

   An alternative direction of the dissertation is to build upon current techniques for automated complexity analysis, with the goal of analyzing open code and/or code with complex data structures. This direction will also build on the work of the VeriFIT group, as well as Dr. F. Zuleger from TU Wien and Dr. M. Sinn from the University of Applied Sciences St. Pölten.

   The dissertation will be carried out in collaboration with the VeriFIT team at the Faculty of Information Technology, Brno University of Technology, which focuses on program verification techniques for programs with complex data structures (in particular Dr. L. Holík, Prof. T. Vojnar, Ing. T. Dacík, Ing. V. Šoková). With a responsible approach and high-quality results, there is an opportunity to participate in research grant projects, including international ones. There is also the possibility of close collaboration with various international partners of VeriFIT: TU Wien, Austria (Dr. F. Zuleger); Verimag, Grenoble, France (Dr. R. Iosif); IRIF, Paris, France (Prof. A. Bouajjani, Dr. M. Sighireanu).

   Within this topic, the student may also actively participate in various grant projects conducted within the VeriFIT group.

   Supervisor: Rogalewicz Adam, doc. Mgr., Ph.D.

4. Automated Techniques for Requirements-Based Test Assessment

   Evaluation of the completeness of test suites with respect to software requirements represents a fundamental challenge in the verification of critical systems, particularly in regulated domains. Current practice is largely based on manual assessment, which is time-consuming and error-prone. This dissertation will focus on the research of formal and automated methods for analyzing requirement coverage by test cases, with an emphasis on the semantic aspects of requirements and their implicit behavior.

   The goal of the dissertation is to propose a formal framework for representing requirements and test cases, and to define objective coverage criteria derived from the semantics of requirements (e.g., combinations of conditions, temporal constraints, sequences of actions, or state transitions). On this basis, methods for computing the coverage level of a test suite and for identifying uncovered semantic cases will be investigated.

   The research will focus in particular on the following areas:

   • design and systematic comparison of formal models suitable for representing different classes of requirements and testing artifacts, including analysis of their expressive power and analyzability;
   • formalization of the notion of requirement coverage and derivation of corresponding coverage obligations from the formal semantics of requirements;
   • design of algorithms for analyzing test suite coverage over formal representations, including identification of uncovered cases and their linkage to specific parts of the requirements;
   • investigation of methods for transforming informal requirements and test cases into formal models, including the use of natural language processing techniques and large language models, with emphasis on the reliability and limitations of these approaches;
   • experimental evaluation of the proposed methods on realistic datasets, including assessment of accuracy, completeness, and scalability.

   The dissertation will also include the design and implementation of prototype tools supporting selected aspects of the proposed approach, along with their validation on examples from embedded and avionics systems. Emphasis will be placed on integrating formal methods, software testing, and automated analysis techniques, with the aim of contributing to systematic and repeatable evaluation of test suite quality.

   Supervisor: Rogalewicz Adam, doc. Mgr., Ph.D.

5. Cognitive Load Mitigation in Human-Robot Collaboration using Mixed Reality

   The objective of this thesis is to investigate the cognitive load and stress levels experienced by workers during direct collaboration with robotic systems. The research utilises Mixed Reality (MR) technologies and Digital Twins to perform an in-depth analysis of factors affecting the human psyche in automated environments. The primary task is to identify the causes of stress across various industrial scenarios and propose methods to prevent or mitigate these negative effects through an adaptive interface.

   The work bridges empirical testing on real robotic workstations with their digital twins, allowing for the safe simulation and optimisation of critical situations. To quantify workload, biometric signals (e.g., heart rate, eye-tracking, or brain activity) will be used to provide feedback for dynamically adjusting the MR environment. By leveraging augmented reality for navigation and diminished reality to filter out environmental distractions, the research aims to create a workspace that adapts in real-time to the operator's current needs and cognitive capacity.

   The contribution of this research lies in defining standards for ergonomic and safe human-robot interaction in the spirit of Industry 5.0. The results will provide the industry with procedures for deploying advanced technologies that increase production efficiency while also protecting employees' mental health and long-term well-being. This approach is essential for the future integration of complex robotics into standard workflows without risking cognitive overload.

   Supervisor: Beran Vítězslav, doc. Ing., Ph.D.

6. Computer-Aided Creativity

   The aim of this dissertation is to conduct research in the field of generative artificial intelligence—whether it involves diffusion and adversarial models for video generation, generative text models for story creation, the automatic generation of computer code and music, the representation of knowledge in physics and chemistry, the promotion of scientific creativity, or a combination of all these approaches. The work will focus on addressing issues related to human interaction with generated intermediate results, the natural labeling of individual parts and concepts so that it is possible to build upon ongoing results, and the development of methods for modifying datasets and learning procedures to address societal issues associated with the creative models being developed—questions of model fairness, bias, and the incorporation of concepts of so-called responsible artificial intelligence.

   Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

7. Detection and Identification of Cryptocurrency Users from Network Traffic

   Motivation:

   Cryptocurrencies are generally used for the nearly instantaneous (i.e., until recorded on the blockchain), decentralized (i.e., without any regulatory authority), and (pseudo)anonymous transfer of cryptoassets (whether coins or tokens) between participants. These and many other characteristics stemming from the inherent use of cryptography are also the reason why cryptocurrencies are an integral part of cybercriminal activities (e.g., extortion via ransomware, purchases/sales on dark markets, investment scams).

   The majority of cryptocurrencies operate as a peer-to-peer network of nodes that mutually update and maintain a replica of the blockchain. Nodes communicate with each other using standardized protocols (e.g., the Bitcoin protocol [1]). Based on messages sent via the protocol, one can infer the state of a node (e.g., the creation of a new transaction) and also collect related identifiers (e.g., IP address, cryptocurrency address, or transaction hash). Furthermore, there are other overlay networks built on top of existing cryptocurrency networks (e.g., the Lightning Network [2]), where the same principles can be applied to interconnect information from them (similar to miner detection [3]), with the ultimate goal being the deanonymization of cryptocurrency users and their addresses. This dissertation could thus make a significant contribution to law enforcement agencies in identifying users of cryptocurrency addresses and linking their identities in the blockchain and network worlds.

   Objective:

   The aim of this work is to develop and validate new methods for analyzing cryptocurrency-related network traffic in order to deanonymize users and their associated identifiers.

   Expected phases of the project:

   1. Research on the use of cryptocurrencies in cybercrime in collaboration with law enforcement agencies.
   2. Analysis of existing approaches to the detection and analysis of cryptocurrency traffic and the use of this data to identify cryptocurrency users.
   3. Implementation of new approaches, deployment, and subsequent verification of their functionality in practice.
   4. Evaluation of the results achieved.

   Project:

   The work will thematically build on the results of earlier projects such as TARZAN and MIXER. The research efforts and expected outputs will be part of the COW (https://www.fit.vut.cz/research/project/c36822/) and COP IDEA (https://www.fit.vut.cz/research/project/c38120/) projects. Funding is primarily expected from the COW project.

   References:

   [1] P2P Network. In: Bitcoin Developer Guide [online]. Bitcoin.org [accessed 2026-04-07]. Available from: https://developer.bitcoin.org/devguide/p2p_network.html

   [2] POON, J. and DRYJA, T. The Bitcoin Lightning Network: Scalable Off-Chain Instant Payments [online]. Draft v0.5.9.2. 2016-01-14 [accessed 2026-04-07]. Available from: https://lightning.network/lightning-network-paper.pdf

   [3] VESELÝ, V.; ŽÁDNÍK, M. How to detect cryptocurrency miners? By traffic forensics!. Digital Investigation, 2019, vol. 31, no. 31, p. 1-14. ISSN: 1742-2876.

   Supervisor: Veselý Vladimír, Ing., Ph.D.

8. Detection and Prevention of Deepfake Attacks on Images of Human Faces

   The student will study techniques used in deepfake attacks on images and videos of human faces. The doctoral study is linked to the TrustedFace project (in collaboration with Innovatrics). The goal is to investigate existing deepfake attacks on images of human faces aimed at impersonating identities, quantify their severity, and propose possible defenses – both for detecting attacks and preventing them. The student will need to gain in-depth familiarity with available datasets of human faces, techniques for face identification and verification in images and video, and deepfake attack techniques. The ultimate goal of the study and doctoral work is to develop defenses against deepfake attacks and to formulate recommendations for systems that utilize face-based identification and verification.

   Supervisor: Herout Adam, prof. Ing., Ph.D.

9. Efficient finite automata for automated reasoning

   We will develop techniques for working with finite automata with applications in automatic reasoning, verification, and decision procedures for logics. Although automata are a fundamental concept of computer science with many applications and strong theoretical foundations, their practical use remains an active and challenging research area.

   The field is strongly motivated by applications in verification and testing, regular pattern search, artificial intelligence and inference, SMT-solving, system design, and automated synthesis. For example, the most efficient algorithms for regular pattern matching are automata-based, yet extending them to features such as backreferences, repetition counters, or look-arounds remains open. String solving has applications in program verification, particularly in security analysis of web applications (e.g., SQL injection, XSS), cloud access policies, and models of critical systems. Another challenge is efficient decision procedures for integer arithmetic and related logics. Regular model checking enables verification of infinite-state systems such as programs with dynamic data structures or communication protocols. Automated techniques can also support program synthesis (e.g., interfaces or drivers).

   These areas raise fundamental theoretical questions about decidability and complexity. For instance, how to model backreferences using automata? What classes of string-manipulating or data-structure programs are decidable, and at what cost?

   From a practical perspective, key challenges include efficient implementation of automata algorithms and mitigating state explosion. This involves compact representations via minimization, abstraction, approximation, and extensions such as alternating, symbolic, counter, or register automata. Their theoretical properties and efficient implementation for real-world problems remain open research directions.

   The supervision team includes Doc. Holík (VeriFIT), Dr. Lengál, Doc. Rogalewicz, Doc. Češka, and Prof. Vojnar, doctor Smrčka. The group is internationally recognized, with extensive publications, software tools, awards, and collaborations (e.g., Microsoft Research, Academia Sinica, Uppsala University). We also collaborate with industry (Red Hat, Honeywell). PhD graduates have strong prospects in both academia and industry, and students are supported through successful grant funding.

   Supervisor: Holík Lukáš, doc. Mgr., Ph.D.

10. Efficient HW accelerators for DNN inference

    Artificial convolutional and transformer neural networks (NNs) have recently been widely used in computationally intensive tasks such as classification, prediction, and recognition. The aim of this dissertation will be to propose methods for optimizing NN inference on embedded devices. The task will be to reduce the energy consumption of the output using advanced techniques, such as finding suitable data and weights representations, network architectures, optimizing computation scheduling on existing platforms, or implementing parts of the NN computation in-house. This research falls within the scope of topics addressed by the EvoAI Hardware research group.

    Supervisor: Mrázek Vojtěch, doc. Ing., Ph.D.

11. Efficient Technology for Dealing with Logics and Automata (Not Only) for Formal Analysis and Verification

    Different types of logics and automata are among the most fundamental objects studied and applied in computer science for decades. Nevertheless, there are many unsatisfactorily solved problems in this area, and new, exciting problems related to ever new applications of logics and automata are constantly emerging (e.g., in the formal verification of finite and infinite state systems with various advanced control or data structures, in decision procedures, in program or hardware synthesis, in quantum program verification, or even in methods for efficient search in various types of data or network traffic).

    The subject of the dissertation will primarily be the development of the state of the art in efficient work with various logics (e.g. over pointer structures, strings, various arithmetic, temporal logics, etc.). To this end, approaches based on different types of decision diagrams, automata, but also e.g. approaches based on the existence of a model of bounded size or on efficient reductions between different types of logical theories will be investigated. In connection with this, methods for efficient work with decision diagrams and different types of automata (automata over words, trees, infinite words, automata with counters, registers, etc.) will be developed. The work will include theoretical research as well as prototype implementation of the proposed techniques and their experimental evaluation.

    The work will be solved in collaboration with the VeriFIT team working on the development of techniques for working with logics and automata and their applications at FIT BUT. There is also the possibility of close cooperation with various foreign partners of VeriFIT: Academia Sinica, Taiwan (Prof. Y. Chen); TU Vienna, Austria (Assoc. F. Zuleger); LSV, ENS Paris-Saclay (Assoc. M. Sighireanu); IRIF, Paris, France (Assoc. A. Bouajjani, Assoc. P. Habermehl, Assoc. C. Enea), Verimag, Grenoble, France (Assoc. R. Iosif); Uppsala University, Sweden (prof. P.A. Abdulla, prof. B. Jonsson); or RPTU, Kaiserslautern, Germany (prof. A.W. Lin).

    Supervisor: Lengál Ondřej, doc. Ing., Ph.D.

12. Embedded Systems for Video/Signal Processing

    The topic focuses embedded image, video and/or signal processing. Its main goal is to research capabilities of "smart" and "small" units that have such features that allow for their applications requiring smyll, hidden, distributed, low power, mechanically or climatically stressed systems suitable of processing of some signal input. Exploitation of such systems is perspective and wide and also client/server and/or cloud systems. The units themselves can be based on CPU/DSP/GPU/AI accelerator, programmable hardware, or their combination. Smart cameras can be considered as well. Applications of interest include:

    • classification of images or objects using machine learning (AI) using traditional methods or through deep convolution networks neural network or similar approaches (e.g. for industrial quality inspection, etc.),
    • parallel analysis of signal(s) and video (e.g. for robust detection of occurrence of object in industrial or surveillance applications),
    • modern algorithms of video, image, and/or signal exploiting "client/server" or "cloud" (with focus on the technlogy) suitable e.g. for mobile technology and/or embedded systems,
    • other similar topics can be individually consulted and considered.

    A possibility exists in collaboration on grant projects, especially the newly submitted TAČR, HE, ECSEL ones (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

13. Energy-aware Embedded Inteligence

    The aim of this dissertation is to investigate models of embedded intelligence that explicitly account for the energy consumption of specific operations and optimise their performance based on specific constraints imposed by individual devices or the entire system. It will also include implementing selected models on suitable hardware for international projects in which the supervisor is involved.

    Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

14. Evaluating Vulnerabilities and Capabilities of Large Language Models

    Despite the impressive performance of large language models (LLMs), these models are still not well understood. A lot of effort is dedicated to evaluating the capabilities of LLMs and creating guardrails to prevent potentially harmful behaviours. However, small prompt variations still significantly alter the behaviour of LLMs, they often fail in tasks that are similar to the ones seen in training data but contain a small change, or they can be easily used to generate confident misinformation. As the large language models increasingly mediate access to information and are increasingly used by expert and non-expert users, it is essential to develop rigorous methods for evaluating their true capabilities and identifying the sources of their vulnerabilities.

    The goal of this research is to build a deeper and more reliable understanding of the LLM behaviour. The first possible research direction involves designing a benchmark that can accurately measure both advanced capabilities (such as reasoning, planning, understanding, math or multilingual capabilities) and common failure modes and vulnerabilities (including hallucination, brittleness, tendency or ease of generating problematic content such as misinformation). The second possible research direction leverages mechanistic interpretability to study the internal structures that drive the LLM behaviours. These tools will be used to investigate why specific vulnerabilities arise, how the capabilities arise, and whether targeted interventions can reduce failures while maintaining core capabilities.

    By integrating robust behavioural benchmarking with mechanistic insights, the project aims to produce a more principled understanding of how LLMs work and how they can be made safer and more reliable. The application domain includes (but are not limited to) low-resource tasks, multilingual understanding, misinformation and other problematic behaviour detection (narrative detection) or even LLM user simulation.

    Relevant publications:

    • Vykopal, I., Pikuliak, M., Srba, I., Moro, R., Macko, D., and Bielikova, M., 2023. 2024. Disinformation Capabilities of Large Language Models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14830–14847, Bangkok, Thailand. Association for Computational Linguistics. https://aclanthology.org/2024.acl-long.793/
    • Zugecova, A., Macko, D., Srba, I., Moro, R., Kopal, J., Marcinčinová, K., and Mesarčík, M., 2025. Evaluation of LLM Vulnerabilities to Being Misused for Personalized Disinformation Generation. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 780–797, Vienna, Austria. Association for Computational Linguistics. https://aclanthology.org/2025.acl-long.38/

    The research will be performed at the Kempelen Institute of Intelligent Technologies (KInIT, https://kinit.sk) in Bratislava in cooperation with researchers from highly respected research units. A combined (external) form of study and full employment at KInIT is expected.

    Supervisor: Bieliková Mária, prof. Ing., Ph.D.

15. Evolutionary machine learning for automated digital circuit design

    .

    Supervisor: Sekanina Lukáš, prof. Ing., Ph.D.

16. Explainable Artificial Intelligence

    Explainable AI (XAI) centres on enhancing trust, interpretability, and usability of complex models. Key areas include creating human-centred explanations, establishing objective evaluation metrics for, and balancing accuracy with interpretability. Recently, the research has shifted toward interactive explanations, evaluating how AI models handle data and bias, transparently communicating result uncertainty, and evaluating the quality of the generated explanation.

    The research will focus on human-centred and cognitive aspects, aligning AI explanations with human psychology and cognitive abilities, highlighting differences between human-human and AI-human explanations, developing interactive, dialogue-based explanations that allow users to ask "what-if" questions and explore model behaviours, and objectively measuring the quality, understandability, and effectiveness of an explanation for a user. The initial application domain will correspond to fact-checking in the context of the fight against desinformation.

    Supervisor: Smrž Pavel, doc. RNDr., Ph.D.

17. Formal Methods in Design and Analysis of Quantum Programs

    Development in the field of quantum computers is moving forward unstoppably. However, for the effective use of quantum hardware, it is necessary to have the right support at the software level, i.e., quantum algorithms for solving problems, as well as tools for compiling, optimizing and synthesizing quantum programs, mapping them to a specific quantum technology, mapping them to a fault tolerant protocol, analyzing, simulating, and debugging quantum programs, etc.

    The subject of the dissertation will primarily be the development of current knowledge in the field of formal methods in quantum program design. The main focus here will be on the use of automata theory, automatic reasoning, and decision diagrams. In connection with this, methods for effective work with decision diagrams and various types of automata (automata over words, trees, infinite words, automata with counters, registers, etc.) will also be developed. The work will include both theoretical research and prototype implementation of the proposed techniques and their experimental evaluation.

    The work will be carried out in collaboration with the VeriFIT team at FIT BUT, which develops techniques for working with logics and automata and their applications. There is also the possibility of close cooperation with various foreign VeriFIT partners: Academia Sinica, Taiwan (prof. Y. Chen); Uppsala University, Sweden (prof. P.A. Abdulla, prof. R. S. Thinniyam); or Leiden Institute for Advanced Computer Science, Leiden, the Netherlands (prof. A. Laarman).

    Supervisor: Lengál Ondřej, doc. Ing., Ph.D.

18. Genetic programming and LLM for automated circuit and program synthesis

    .

    Supervisor: Sekanina Lukáš, prof. Ing., Ph.D.

19. Hardware architectures and methods for mapping neural networks onto FPGAs

    In recent years, neural networks (NNs) have achieved breakthrough results in tasks such as image recognition, natural language processing, object detection, and predictive control. Growing interest in deploying NNs in applications with strict response requirements—such as autonomous vehicles, industrial robotics, stock trading, or network traffic monitoring—is creating pressure to reduce inference latency to the range of hundreds of nanoseconds to microseconds.

    While general-purpose computing platforms (CPUs, GPUs) offer high flexibility, they face physical limitations imposed by data transfer between memory and the processor, as well as high power consumption. In contrast, FPGAs (Field-Programmable Gate Arrays) allow data paths, parallelism, and computational accuracy to be fully tailored to a specific network architecture. Thanks to their reconfigurability, deterministic latency, and low power consumption, FPGAs represent a promising platform for NN acceleration, particularly in edge computing and embedded systems.

    Nevertheless, key research questions remain open: how to efficiently map various network topologies onto a limited chip area, how to minimize latency while maintaining acceptable accuracy, and how to automate and generalize this process for different families of NNs and FPGAs.

    Objectives of the Dissertation

    1. Analysis of the Current State – A systematic review of existing approaches to mapping NNs onto FPGAs, identifying their limitations in terms of latency, throughput, and energy efficiency.
    2. Design of new hardware architectures – Research and design of specialized processing elements and memory hierarchies optimized for low-latency inference of convolutional (CNN), recurrent (RNN), and transformer architectures.
    3. FPGA-Targeted Network Quantization and Pruning Methods – Research into bit-width reduction techniques (mixed-precision quantization), structural pruning, and model compression tailored to the specifics of FPGA architecture, with the goal of minimizing computational and memory requirements while maintaining accuracy.
    4. Automated mapping (HLS and others) – Design and verification of methodologies or tools for the automatic generation of FPGA implementations from high-level NN descriptions, with latency optimization as the primary criterion (extensions of HLS approaches, e.g., Vitis HLS, or direct synthesis at the RTL level).
    5. Experimental validation – Implementation and measurement of the proposed architectures on real FPGA hardware (e.g., Xilinx/AMD Ultrascale+, Intel/Altera Agilex) and comparison with reference solutions (GPUs, existing FPGA frameworks such as hls4ml, finn).
    6. Generalization and transferability – Formulation of general principles and recommendations for the design of low-latency FPGA NN accelerators applicable across various domains.

    Expected benefits

    • New insights into the relationship between NN architecture, mapping strategy, and achieved inference latency on FPGAs.
    • Design and verification of hardware architectures that surpass the current state of the art in terms of latency.
    • Publication of results in peer-reviewed international journals and at conferences (FPL, FCCM, DAC, DATE, ICCAD).

    Supervisor: Martínek Tomáš, doc. Ing., Ph.D.

20. High-Performance Traffic Shaping for Multi-Gigabit Edge Networks

    The transition of residential and access network infrastructure to multi-gigabit plans for end customers has invalidated many of the assumptions on which the network traffic quality of service subsystem within the Linux kernel is based. Traditional algorithms, such as Hierarchical Token Bucket, suffer from a significant performance degradation at speeds of several gigabits due to the need to use global locks and cache locality violations.

    This dissertation focuses on the optimization and design of new algorithms and architectures to meet network service level agreement (SLA) parameters, even in a hierarchical network topology. The thesis aims to explore new mechanisms and lock-free algorithms with distributed state management at the individual processor level. The research will examine trade-offs between flexibility and efficiency in the Linux kernel so that the proposed algorithms can meet the required parameters even at speeds of several gigabits per second while ensuring the necessary level of quality of service.

    Supervisor: Matoušek Petr, doc. Ing., Ph.D., M.A.

21. Homographic Navigation – augmented reality for capturing precise images of objects

    The aim of this work will be to develop the technique of "homographic navigation," a new concept being developed under the guidance of the doctoral program supervisor. The PhD student will develop computer vision algorithms and related machine learning techniques. Model training for homographic navigation must be minimally supervised. Both the learning algorithms and the inference and use in the application must be fast and resource-efficient. The project also involves prototyping and developing an augmented reality application that will be used for data collection, user testing, and other related tasks. Homographic navigation technology is intended to have immediate industrial applications, so the doctoral study will also include the development of all essential techniques through to the production phase.

    Supervisor: Herout Adam, prof. Ing., Ph.D.

22. Human in the loop (HIL) in the adaptation of AI systems

    This PhD topic is driven by the needs of users for which the privacy of data is crucial and concentrates on improving AI (especially speech recognition) system in case the user is able to identify and correct systems errors. The topic includes proper evaluations of HIL systems, selection of data to be proposed for correction and actual fine-tuning / adaptation techniques working with large pre-trained models. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

23. Making Machine Learning Alive

    Towards "Living" Machine

    Bio-Inspired Recurrent Architectures: Integrating Top-Down Feedback  for Adaptive Inference in Non-Stationary Environments.

    9Problem Statement

    Unlike biological systems, which continue to learn and adapt during the inference process, standard ML models cannot self-correct or reconfigure their internal processing streams when faced with novel or shifting data distributions. This lack of dynamic inference leads to catastrophic failure in changing environments.

    Research Objectives

    The goal of this research is to move beyond "dead" feedforward architectures by implementing three key biological principles:

    • Reciprocal Recurrence: Replace unidirectional chains with parallel processing streams and massive top-down feedback loops.
    • Active Inference Learning: Develop mechanisms that allow the model to evaluate the quality of its own extracted information in real-time.
    • Gating: Implement weighting of lower-level neuron outputs, controlled by higher-level feedback to filter irrelevant data and focus on contextually salient features.

    Expected Contribution

    This research aims to deliver a blueprint for a new class of Self-Correcting Artificial Intelligence. By mimicking the "dynamic and alive" nature of biological perceptual systems, we can create AI that doesn't just "predict," but actively "perceives" and adjusts its strategies in real-time—essentially turning inference from a static snapshot into a continuous, adaptive process.

    Supervisor: Heřmanský Hynek, prof. Ing., Dr. Eng.

24. Methods for knowledge extraction and representation from heterogeneous sources to improve the performance and interpretability of AI systems

    Currently, there is a dynamic development of artificial intelligence methods, particularly large language models, which enable advanced processing and interpretation of extensive unstructured data. However, efficient knowledge mining from heterogeneous data sources and their transformation into a form suitable for training and improving AI systems remains an open problem. The quality and structure of input knowledge fundamentally influence the resulting accuracy, robustness, and interpretability of these systems.

    With the increasing complexity of data and requirements for automated decision-making, there is a growing need to develop methods that allow for systematic extraction, representation, and utilization of knowledge across various domains. This issue has both significant theoretical contribution and broad application potential in the field of intelligent agents and decision support systems.
    The aim of the doctoral thesis is to design, implement, and experimentally verify new approaches to knowledge mining, representation, and utilization from heterogeneous data sources to improve the performance of AI systems based on language models.

    1. Analysis of the current state of knowledge
       • Systematic review of knowledge mining methods and their application in AI
       • Identification of the limits of existing approaches when working with unstructured and multimodal data
    2. Design of knowledge mining methods
       • Development of new approaches for extracting relevant knowledge from diverse data sources (text, tables, documents, logs)
       • Design of methods for integrating structured and unstructured data
    3. Representation and integration of knowledge into AI systems
       • Design of suitable knowledge representations (e.g., knowledge graphs, embedding representations)
       • Optimization of knowledge utilization in language models (e.g., retrieval-augmented generation, fine-tuning)
    4. Experimental verification of the proposed approaches
       • Implementation of the proposed methods
       • Testing on real-world and publicly available datasets
       • Comparison with existing methods using appropriate metrics
    5. Verification of transferability and generalizability
       • Application of the proposed methods to various domains and use cases
       • Evaluation of the robustness and adaptability of the approaches
    6. Publication and application outputs
       • Publication of results in scientific journals and at conferences
       • Creation of prototype tools or frameworks

    At the same time, the thesis will be conceived generally so that the results are transferable to other areas where working with knowledge and data is key, such as enterprise information systems, decision support systems, autonomous AI agents, or applications in industry, healthcare, or public administration.

    Supervisor: Burget Radek, doc. Ing., Ph.D.

25. Modeling Latent Representations of Speech Attributes Using Generative Models

    This work focuses on the study and design of generative neural models for learning latent representations of speech signals, with an emphasis on the effective capture and disentanglement of individual speech attributes such as speaker identity, emotion, prosody, and articulation. The goal is to develop models that enable the disentanglement of these attributes in the latent space and their subsequent controllable manipulation during speech synthesis. The research will include an analysis of existing approaches (e.g., variational autoencoders, generative adversarial networks, and diffusion models), the design of novel methods, and their experimental evaluation on real-world speech data. The expected outcome is an improvement in the quality and flexibility of generated speech signals, as well as a deeper understanding of speech representation in generative models.

    Supervisor: Burget Lukáš, doc. Ing., Ph.D.

26. Modern Algorithms of Computer Graphics

    The topic concerns algorithms of computer graphics and image synthesis. Its main goal is to research new algorithms so that their features and application possibilities are better understood so that they are improved or newly created. If suitable, it is possible to work on various platforms, includeing parallel CPUs, such as x86/64, ARM, Xeon PHI, GPU, etc. or other cores in CUDA, OpenCl, VHDL, etc. Algorithms of interest include:

    • rendering using selected computer graphics methods (such as ray tracing, photon mapping, direct rendering of "point clouds", etc.),
    • modeling of scenes and redering using artificial intelligence, including image synthesis using neural netowrks (especially CNNs),
    • processing and rendering of "lightfield" images, their acquisition, or possibly compression, reconstruction of 3D scenes from images and/or video, eventually also fusing with other sensors, such as LIDAR or RADAR,
    • modern algorithms of geometry suitable for applications in cpmputer graphics and perhaps also 3D printing,
    • emerging algorithms of 3D synthesis, holography, wavelet transform, frequency transform, etc.

    After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

    Collaboration on grant projects, such as TACR, HE, ECSEL (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

27. Multimodal Analysis of Forensic Evidence Using Artificial Intelligence Methods

    This dissertation will focus on the development of advanced methods for the analysis of forensic evidence using multimodal data, particularly image and textual information. The rapid advancement of artificial intelligence methods—especially deep neural networks and large pre-trained models—significantly expands the possibilities of automated processing and interpretation of forensic materials. This progress contributes to increased efficiency and objectivity in forensic examination, including applications within the Customs Administration and other security agencies. A key challenge lies in the robust analysis of unconventional and limited datasets.
    The research will address multimodal analysis of forensic evidence with an emphasis on practical applicability, specifically:
    (1) the design of new methods for analyzing unconventional image data, such as X-ray images from inspection systems used by the Customs Administration;
    (2) the development of methods for data augmentation and synthetic dataset expansion to improve the robustness and generalization of models in detecting risk-related objects; and
    (3) the integration of computer vision methods with natural language processing techniques for joint evaluation of visual evidence and related textual documentation.
    The overall aim of the dissertation is to propose and validate a robust multimodal approach to forensic evidence analysis applicable in practice within the Customs Administration and other security agencies.

    Supervisor: Hanáček Petr, doc. Dr. Ing.

28. Network configuration generated using LLM models

    Large LLM language models are now applied in various areas where they perform common and repetitive activities. LLM support can also be used in network design and configuration of network devices, where they can generate a configuration in the domain language of a given manufacturer (e.g. Cisco IOS, Juniper OS, etc.) based on instructions in the recommended language. However, configuration generation requires accuracy and the ability to detect possible errors that arise from inaccuracies in input, hallucinations of LLM models, etc.

    The assigned topic develops the principle called Intent-based Networking, where we will use LLM language models and a system of cooperating agents to find a procedure for implementing the desired network behavior and generate the corresponding configurations for specific network devices. In order to be able to deploy the given procedure in real-world networks, it is necessary to design a method for verifying the generated configuration by combining static and dynamic verification, which will verify whether the given configuration meets the required properties.  

    Supervisor: Matoušek Petr, doc. Ing., Ph.D., M.A.

29. Neurosymbolic methods for safe agent control

    The subject of the dissertation is the development of theoretical foundations and algorithms for safe agent control using neurosymbolic AI, which combines deep reinforcement learning methods and symbolic formal methods. Emphasis will be placed on approaches that guarantee safe and explainable solutions. The dissertation will focus on the synthesis of decision trees and finite-state controllers, as well as on monitoring methods and runtime verification for agent systems. The dissertation will also examine model-learning methods and model abstractions enabling the utilization of formal methods. The work will focus on the use of these methods in the field of agent control in environments with probabilistic dynamics and partial observability, and on a thorough evaluation of the applicability of these methods. The results of this work will contribute to the field of autonomous systems.

    Research on neurosymbolic control methods is currently receiving considerable attention in the fields of artificial intelligence and formal methods, as evidenced by the focus of a number of top conferences (e.g. AAAI, AAMAS, CAV, or TACAS).

    The work will be carried out in cooperation with the VeriFIT team under the leadership of doc. M. Češka. Furthermore, close cooperation is expected with the group of Prof. N. Jansen (ERC Starting grant investigator), Bochum University, Germany, Prof. D. Parker, Oxford University, UK and Prof. S. Junges Radboud University Nijmegen, Netherlands.

    In case of a responsible approach and high-quality results, there is a possibility of participation in grant projects (e.g. Czech Science Foundation or Horizon Europe projects).

    Supervisor: Češka Milan, doc. RNDr., Ph.D.

30. New approaches to the numerical solution of partial differential equations

    Many technical problems are described by the partial differential equations. These equations are very difficult to solve analytically and many approaches exist that allow for the numerical approximation of the solution. The aim of this PhD thesis is to study the existing approaches to the numerical solution of the partial differential equations and develop new approaches to their solution. The student will mostly concentrate on the solution of the system in time domain, becoming acquainted with the wide range of methods used for the numerical solution of the differential equations.

    Supervisor: Jaroš Jiří, prof. Ing., Ph.D.

31. Non-Traditional Visal Computing Algorithms

    The topic of the work includes algorithms of comptuer graphics, image, and video, i.e. "Visual Computing", such as HDR (High Dynamic Range) image, multispectral image, stereoimage, possibly also image augmented with material features, temperature, etc. The goal is to understand the features and possibilites but also applications, to analyze the algorithms in-depth, prepare new and improve them. The possible algorithms include:
    
    

    • Acquisition of HDR image and video, its compression, assembling from several standard images, videos, etc.,
    • acquisition, processing, and visualization of mutispectral images (images with more than three spectral components),
    • processing of stereo images and other images containing the depth information using both structural light illumination or no structural light,
    • algorithms suitable for mobile technology and/or embedded systems focusing the above types of iamges, or images acquired from drones,
    • metods of visualization of HDR images and video, tone-mapping, real-time tone mapping,
    • algorithms and applications exploiting Wavelet, frequency, and/or other similar transformations.

     

    After mutual agreement, individually selected algorithms can be considered as well as soon as they do belong to the general topic.

    Collaboration on grant projects, such as TACR, MPO, HE, ECSEL (employment expected).

    Supervisor: Zemčík Pavel, prof. Dr. Ing., dr. h. c.

32. Performance of voice deepfake detection by humans and machines

    The work will start with getting familiar with the basics of the problem of voice deep fake detection (DFD), terminology, available techniques, data and challenges (AVSpoof, WildSpoof), with the history and state of the art techniques and tools for speaker recognition (wespeaker toolkit), with state of the art techniques and tools for personalized text to speech (pTTS) synthesis and voice conversion. The topic will then concentrate on the evaluation of human and machine performance in DFD, while concentrating on the motivation aspect of human subjects to simulate true attacks on people. The PhD will then progress in both (1) technical (advances in deepfake generation and detection technology) and (2) human-aspects. The work counts on collaboration with specialists from psychology and sociology. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

33. Performance of voice deepfake detection by humans and machines

    The work will start with getting familiar with the basics of the problem of voice deep fake detection (DFD), terminology, available techniques, data and challenges (AVSpoof, WildSpoof), with the history and state of the art techniques and tools for speaker recognition (wespeaker toolkit), with state of the art techniques and tools for personalized text to speech (pTTS) synthesis and voice conversion. The topic will then concentrate on the evaluation of human and machine performance in DFD, while concentrating on the motivation aspect of human subjects to simulate true attacks on people. The PhD will then progress in both (1) technical (advances in deepfake generation and detection technology) and (2) human-aspects. The work counts on collaboration with specialists from psychology and sociology. 

    Supervisor: Černocký Jan, prof. Dr. Ing.

34. Self-evaluation Machine Learning

    Research Gap: The "Blind" Inference Problem

    Current AI suffers from "silent failure." When a model encounters data outside its training distribution, it still provides an output—often with high overconfidence—because it lacks a biological Self-Evaluation Module. It cannot "feel" its own confusion. Biological systems survive because they evaluate the quality of extracted information during the act of perceiving, allowing them to switch strategies instantly.

    Research Focus: The SEM

    This research proposes the development of an independent Self-Evaluation Module (SEM) that sits atop standard processing layers. Its primary functions include:

    • Confidence Synthesis: Moving beyond simple "Softmax" probability to multi-factor reliability metrics (e.g., entropy analysis, reconstruction error, or latent space stability).
    • The "Surprise" Trigger: Detecting when new data contradicts internal world models (novelty detection) to trigger an "adaptation phase" during a single forward pass.
    • Efferent Control Gating: Using the evaluation score to send top-down signals that "gate" (turn on/off) specific lower-level neurons, effectively reconfiguring the model's focus in real-time.

    Key Research Questions

    • How can a model calculate the "quality" of its own inference without access to ground-truth labels?
    • How do we prevent "feedback loops" where a self-evaluating model becomes stuck in a recursive state of doubt?

    Impact

    By perfecting the Self-Evaluation Module, we move toward AI that knows when it doesn't know. This is critical for safety-critical applications like autonomous driving or medical diagnostics, where a machine’s ability to flag its own uncertainty is as important as the prediction itself.

    Supervisor: Heřmanský Hynek, prof. Ing., Dr. Eng.

35. Speech recognition and data processing for ATC-pilot communication in aviation

    This PhD topic is focused on speech recognition and data processing for ATC-pilot communication in aviation. It will cover all components of automatic speech recognition (ASR), i.e. data processing, acoustic model, vocabulary (including special aviation terminology), language model as well as interaction with sources of meta-data (radar information, airport). Special attention will be paid to the use of large pre-trained model, code-switching, reliable language identification from short segments and adaptation to the conditions of given airport with crawled data.

    Supervisor: Černocký Jan, prof. Dr. Ing.

36. String constraint solving in applications

    We will develop techniques for string constraint solving, i.e., automated reasoning about logics over strings, with applications in verification, testing, and security. String solving has recently reached a level of maturity that enables precise analysis of string-manipulating programs in real-world systems .

    The student will contribute to advancing the core technology, focusing on improving scalability and expressiveness of string solvers, in particular within the SMT-solving framework. This includes automata-based techniques, integration with other theories (e.g., integers, arrays), and handling advanced string operations such as replacements, concatenation, and string-number conversions.

    A key part of the work will be applying string solving in challenging domains such as security analysis of web applications, analysis of access control policies, and testing of safety-critical systems. These applications require reasoning about complex relations between string values that cannot be handled by traditional testing or static analysis techniques.

    The student will also participate in a joint project with Embraer, focusing on applications of SMT solving, string solving, and related techniques in industrial settings. This includes advanced testing, static analysis, and verification of real-world systems, and may involve a stronger engineering component, including implementation and integration into existing toolchains.

    The research will be conducted within the VeriFIT group (L. Holík, O. Lengál, A. Rogalewicz, M. Češka, T. Vojnar), an internationally recognized team with strong publication record, industrial collaborations, and experience in developing state-of-the-art verification tools. PhD students benefit from close collaboration with leading academic and industrial partners and strong career prospects in both academia and industry.

    Supervisor: Holík Lukáš, doc. Mgr., Ph.D.

37. The impact of the life cycle of large language models on cybersecurity

    In recent years, there has been an increase in the use of neural networks for generating synthetic content, which goes hand in hand with the rise of new cybersecurity challenges. Generative models can have various impacts on cybersecurity, ranging from positive to negative.A significant area is the security of deployment and operation of generative models, primarily large language models.

    The goal of this thesis is to identify problem areas in a selected field of LLM deployment and operation (e.g., model inference attacks, model theft, or information theft) and analyze new trends, approaches, defenses, and their characteristics, impacts, and potential applications. The work should then propose new protection methods based on the analysis and research on the state of security for the selected areas.

    Recommended areas of work focus:
    Attacks and defenses in the side channel domain of large language models
    Attacks and defenses in the inference domain of large language models
    Defences in the area of adversarial attacks on large language models

    Participation in relevant international conferences and publication in peer-reviewed or scientific journals are expected.

    Supervisor: Malinka Kamil, doc. Mgr., Ph.D.

38. Unconventional Wave Propagation Models

    The aim of the dissertation is to develop unconventional models for acoustic wave propagation. The work will focus on advanced acoustic propagators, Gaussian beam models, and decompositions of spectral methods. The goal is to create a hybrid model that uses k-Wave for low frequencies, while employing advanced propagators for high frequencies. This will significantly reduce both spatial and temporal complexity. In addition to designing the algorithm itself, we will also work on the efficient implementation of these models on clusters of GPU workstations.

    Supervisor: Jaroš Jiří, prof. Ing., Ph.D.

39. Visual Geo-Localization and Augmented Reality

    The project deals with geo-localization of images and videos in unknown environments using computer vision and computer graphics methods. The aim is to investigate and develop new image registration techniques (with geo-localized image database or 3D terrain model). The goal is an efficient implementation of proposed methods on mobile devices as well as search for additional applications in the area of image processing, computational photography, and augmented reality.

    • Further information: http://cadik.posvete.cz/locate/
    • Contact: http://cadik.posvete.cz/
    • Cooperation and research visits with leading research labs are possible (Adobe Research, USA, MPII Saarbrücken, Germany, Disney Research Zurich, Switzerland, INRIA Bordeaux, France)

    Supervisor: Čadík Martin, prof. Ing., Ph.D.