News

AI Awards for Professor Hynek Heřmanský: The significance of an extraordinary scientific figure recognized at home as well

On Thursday, December 4, this year's AI Awards were presented at Vzlet in Prague. These are the only professional awards in the field of artificial intelligence in the Czech Republic. Professor Hynek Heřmanský, who has close ties to the Faculty of Information Technology at Brno University of Technology and Brno University of Technology in general, received one of the six awards presented this year: Specifically, for his long-term and extraordinary contribution to the development of speech recognition and machine processing, which is a key discipline in today's artificial intelligence. Without exaggeration, we can describe him as a personality who has contributed to shaping the current form of speech recognition and neural networks. Broadly speaking, his research focuses on the detection, identification, classification, and reconstruction of information from sensory signals. Heřmanský's fundamental influence on the field is confirmed by a number of top international awards, including the ISCA Medal (2013), the IEEE Award (2020), and the status of IEEE Life Fellow.

Heřmanský's personal and professional journey is in itself an incredible story of what a Central European with a passion for modern technology has been able to achieve in recent decades. And this despite historical circumstances. He was born in 1946 in Nové Město na Moravě, graduated in electrical engineering from the Brno University of Technology in 1972, and after various vicissitudes, he managed to travel abroad and study for his doctorate at Tokyo University in Japan. In the 1980s, he was already working in the United States in industrial research for companies such as Panasonic Technologies and U.S. West Advanced Technologies. In the first half of the 1990s, he became a professor at the Oregon Graduate Institute in Portland, where many Czech students (especially from Brno) came to study with him at the beginning of the millennium, for whom it was a turning point in their careers. This was also commented on by Professor Honza Černocký, who leads the "speech" research group at FIT and was also a member of the AI Awards expert jury: "Hynka's lifetime achievement award is also a recognition of the fact that at BUT we can do science at a decent level in this field, that we have talented people at the faculty, and that we have international projects and contacts. The period around 2000, when our entire Brno group gradually passed through Hynka's laboratory in Oregon, was a crucial period. We moved from local conditions to the center of world events in speech processing, which is also called artificial intelligence today." Hynek Heřmanský himself approaches the award with unfeigned modesty and perspective: "Honestly, I'm happy about it. But I have no idea how they found me. I absolutely did not expect the invitation. I receive awards abroad, but at home it's a surprise. I don't think many people outside professional circles know about me here. I don't crave fame; in fact, I would actively resist it."

Winners of individual categories of the AI Awards. Hynek Heřmanský third from the right | Author: Prg.ai, adam&costey

In 2003, Heřmanský moved to Europe when he took up the position of research director at the prestigious IDIAP Research Institute in Martigny, Switzerland. He was also appointed titular professor at the Swiss Federal Institute of Technology EPFL in Lausanne. At the end of 2008, he took another major step – he accepted an offer from Johns Hopkins University in the USA and also became head of the prestigious Center for Language and Speech Processing (CLSP). He headed this center for 12 long years. His team at JHU included dozens of experts and students from around the world. Since 2022, Hynek Heřmanský has been a professor emeritus at Johns Hopkins University. He continues to maintain close ties with Czechia, which is made easier by the fact that he now spends most of the year in Brno. He works as a researcher in the Brno-based Speech@FIT research group, where he helps lead projects and mentor young scientists. His scientific career has thus symbolically come full circle: after five decades abroad, he has returned to the place where he studied at university. And he continues to pass on his experience to the next generation of Czech researchers. We are grateful to him for that!

It should be added that the AI Awards aim to highlight successful stories, personalities, companies, and organizations that have significantly contributed to the development of artificial intelligence in Czechia. In addition to Hynek Heřmanský, this year's awards went to the EDIH CTU center for creating the ADAM chatbot for the audit body of the Ministry of Finance; the Department of Social Work at the Faculty of Arts of Charles University for the deployment of artificial intelligence to reduce the administrative burden on social workers; the multi-robotic systems team of the Faculty of Electrical Engineering of the Czech Technical University and Fly4Future for their success in the Autonomous Drone Racing League competition; the AI for Children initiative with its artificial intelligence curriculum for primary and secondary schools; and finally, the Czech technology company E2B and its solution for the safe development of AI agents. This year's AI Awards were held under the auspices of Czech President Petr Pavel. The AI Awards are organized by the Czech National AI Platform, and the main partner of the event is the innovation agency JIC.

We will return to the life story and scientific journey of Hynek Heřmanský in a longer article in the coming days.

Author: AI Awards Archive

Hynek Heřmanský together with colleagues from the Speech@FIT research group

[img]

The European ISOLDE project aims to strengthen the continent's technological sovereignty. FIT BUT is involved in solutions in the automotive sector.

A major European project with an ambitious goal is approaching its final phase: to transform the European technology scene by creating an open, high-performance, and secure microchip design capacity ecosystem built around the RISC-V microchip architecture. This could reduce Europe's dependence on non-European microchip suppliers. The ISOLDE project has been running since May 2023 and is currently scheduled to end in May 2026. Funded under the Chips Joint Undertaking (Chips JU), a public-private partnership within the pan-European Horizon Europe program, the project brings together a consortium of 37 partners and 2 associated partners from leading European research institutions and industries. Institutions and organizations from nine European countries are represented. These include FIT VUT, represented by the principal investigator for FIT, Prof. Pavel Zemčík, and Ing. Václav Šimek.

RISC-V is an open microchip architecture (processor instruction set) that allows for free customization and sharing of designs. The project's roots date back to the University of California, Berkeley, and the year 2010. The RISC-V architecture allows for the wide use of microchips based on it in embedded systems, mobile phones, automotive, or even IoT devices. Thanks to the open-source nature of the outputs, European companies will be able to develop and improve chips without licensing restrictions, which supports innovation and independence from proprietary technologies.

FIT VUT contributes to the project as part of a Czech consortium that also includes the Czech branch of NXP (which deals with control algorithms) and Codasip (modification of hardware architecture in cooperation with experts from FIT). The consortium focuses on development in the automotive application area, specifically on the functionality and control of various types of engines and motors used, for example, in cars (window lowering, power steering, etc.). Currently, processors based on ARM core variants are predominantly used in the automotive sector, but this is a licensed solution. The aim is therefore to develop our own solution based on RISC-V architecture, which will be open-source. Work is underway to modify the instruction set and, consequently, the hardware solution of the architecture so that motor control algorithms (such as FOC and MPC) can be executed efficiently within the given RISC-V architecture (in terms of the required chip area, power consumption, and execution speed). The end result should be a practical technology demonstrator, e.g., in the form of a development board with FGPA. The RISC-V design will be described in SystemVerilog at the RTF level, adds Václav Šimek. He himself points out that the project also has a specific impact on studies at the faculty: Currently, three bachelor's theses and one master's thesis are being prepared in connection with it.

The goal of the ISOLDE project is to advance the proposed high-performance processors and IP blocks based on the RISC-V architecture to TRL 8 technological readiness level by the time the project is completed. Approximately two years after completion, the first components are expected to be integrated into industrial products. To ensure long-term sustainability and availability, ISOLDE is initiating the creation of an open-source ecosystem for development, verification, and maintenance. Currently, the project has developed methodologies that enable the smooth integration of open-source and proprietary IP blocks and ensure effective interoperability, as well as the first implementations of demonstration solutions across key sectors such as aerospace, automotive, and IoT. In the final phase of the project, efforts will focus on production-ready designs, prototypes based on FPGA technologies, expanded industrial collaboration, and preparation for the deployment of the developed solutions. The progress of the project shows that Europe can take on a key role in the field of open computing innovation.

Author: ISOLDE Project


[img]

Christmas Fir Trees in Pots at FIT: Treat Yourself and Support Nature

The Faculty of Information Technology BUTis joining the traditional Brno charity event Christmas Fir Trees in Pots organized by the Veronica Foundation for the first time this year. Students and staff can purchase a small Christmas tree that won’t end up as waste after the holidays — instead, it can be planted back in the forest in spring.

The sale at FIT will take place on Tuesday, 9 December 2025, from 10:00 to 12:00 (or until 14:00 depending on interest) at the entrance to Building A.

The trees are approximately 40 cm tall seedlings of silver fir, grown in the nurseries of the Masarykův les Křtiny. The price is 169 CZK, and each tree comes with care instructions and information about the joint spring planting, which will take place on 21 March 2026 near the Křtiny Arboretum. Payment can be made in cash or via QR code.

By purchasing a tree, you support Brno-based non-profit organizations—for example, the Veronica Foundation donates part of the proceeds to the Together for Nature fund, which supports community environmental projects across Moravia. The project also includes free fir trees for single parents, disadvantaged families, and seniors. Another goal of the initiative is to help return the silver fir back to Czech forests.

[img]

David Kozák and František Vídeňský from the Institute of Intelligent Systems will defend their doctoral thesis in December

We invite you to attend the defense of the doctoral thesis of Ing. David Kozák from the Department of Intelligent Systems, FIT BUT, which will take place on Wednesday, December 10, 2025, at 10:30 a.m. in meeting room G108 at FIT VUT. The supervisor of the dissertation entitled "New methods of static analysis not only for ahead-of-time compilers of object-oriented languages" is Prof. Tomáš Vojnar.

David Kozák's dissertation deals with the use of static analysis methods in the context of ahead-of-time compilation of object-oriented languages such as Java. Kozák focuses on the possibilities of applying these techniques in automated analysis and, at the same time, in the field of software architecture reconstruction in a microservices environment. David's research work has a specific application level. The tool in which he implemented most of his research is called GraalVM Native Image. It performs ahead-of-time compilation and has been under development for a long time at the Oracle Labs research institute. Kozák is part of the GraalVM team there.

The main goal of the author's work was to streamline the chain of processes connecting points-to static analysis and the subsequent work of the ahead-of-time compiler. Static analysis can be broadly defined as anything that analyzes a program without running it, with the aim of determining the properties of the program or verifying that it works correctly according to a specified definition of correctness. From the outset, this method has been closely linked to compilers. Static analysis applied to the compiler environment must be fast, efficient, and computationally inexpensive. The specific goal of the author's research was to maintain the reliability and accuracy of points-to static analysis, but at the same time to speed it up, i.e., to approximate the strength of another type of static analysis: rapid type. In addition, this had to be achieved in an ahead-of-time compiler environment, in this case working with a low-level representation of the Java programming language called bytecode. As the name suggests, ahead-of-time compilation into machine code takes place before the program is run as a separate process, only once and without the "luxury" of being able to subsequently interpret code that has not been compiled or, if necessary, recompile certain methods at runtime. Therefore, everything that could be executed is compiled, which is computationally demanding by its very nature, and the resulting binary file is usually large. This is where the motivation lies to streamline and speed up static analysis while maintaining its correctness.

The Native Image compiler that Kozák works with uses type points-to analysis working with object types (classes). Kozák used the saturation technique, which can be simplified as a compromise between points-to analysis and rapid-type analysis. Its main idea can be described as follows: We look at which specific implementations of a given type have been created (instantiated) and use only those. We cover the parts of the program that we can analyze accurately with points-to analysis methods. Conversely, if we have too much data available for some parts of the program, we use rapid-type analysis for them. The result is accuracy close to the outputs of points-to analysis and speed very close to that offered by rapid-type analysis.

Another part of Kozák's dissertation and research task was the design of an extension to the Native Image tool's points-to analysis called SkipFlow. This extension now tracks not only objects but also the flow of primitive values, while using predicate edges to prevent the propagation of values from unreachable branches (dead code) of the program. SkipFlow reduced the load on the compiler and offered more accurate analysis at the cost of a slight increase in runtime and with the benefit of subsequent faster work by the compiler itself.

Kozák's research also has a second, perhaps somewhat surprising and atypical field of application for static analysis – the sphere of microservices, or rather understanding the impact of their interdependencies. And this is more the domain of software engineering. Today, applications are developed in the form of a series of smaller microservices. The problem, however, is that it is difficult to track the communication between individual parts and the impact of changes in individual microservices on other services. Kozák again used the Native Image tool, this time to obtain information about each individual service, and then used this to create a holistic view of the entire application. The compiled version of the application can thus be used to reconstruct the high-level architecture and its completely up-to-date visualization, which can be passed on to the software architect. Kozák also participated in the creation of the SAVAT tool, which is capable of analyzing the impact of changes within the time series of individual microservice versions. Bottom line: David Kozák's dissertation is a step towards unifying three previously largely separate areas: compilers, interprocedural static analysis covering the entire program, and software engineering.

When asked about the key results of his research, David Kozák is clear: "I am most proud of SkipFlow, an extension of points-to analysis that tracks primitive values and monitors dependencies between program branches. Why? Because we have achieved the ideal result here: we have created an analysis that is both more accurate and faster. Then I would mention saturation, because by combining two types of analysis and calculating very accurately only where it makes sense, we were able to achieve almost identical results to a more accurate analysis, but in significantly less time." However, when taking stock, David also mentions more general points: "I am satisfied that this is not purely basic research, but that everything is integrated into a real tool, Native Image, and that saturation and SkipFlow are automatically enabled in this tool. This means that anyone who uses Native Image also uses the results of my dissertation. I like to focus my research on what is used in practice. I enjoy contact with industry and the real-world application of technology." David would like to thank his colleagues from the VeriFIT research group for their consultation and support during his research. As for his future work plans, he states that he would like to continue to straddle academic research and industry. His dream goal is to assemble a larger team of people who would participate in research within the GraalVM team, including bachelor's and master's students interested in compilers.

We would like to add that another public defense of a doctoral thesis is planned at our faculty in the same week. The author of the dissertation "Late Binding of Variables in AgentSpeak(L) Interpreters" is Ing. František Vídeňský from the Department of Intelligent Systems, and his supervisor is doc. František Zbořil. The public defense of this dissertation will take place on Friday, December 12, 2025, at 1:00 p.m. in room G108.

We wish both researchers continued success and joy in their field.

[img]

Ondřej Lengál and Kamil Malinka received their appointment decrees as associate professors. Congratulations!

The ceremonial confirmation of the fact, known for some time now, that the Faculty of Information Technology has two new associate professors in its ranks, took place on Wednesday, November 26, in the auditorium of the BUT Rector's Office in Brno. Ondřej Lengál and Kamil Malinka, both from the Department of Intelligent Systems, have added the new academic title to their names.

Kamil Malinka's habilitation lecture, "Protection mechanisms against deepfake-based attacks," delivered in June of this year, summarizes Malinka's long-term research activities, which he also develops within the Security@FIT research group. Malinka focuses on the broader context of cybersecurity in AI. His research is not limited to technical solutions – he is also interested in behavioral aspects, such as user behavior when using selected security tools or the education of future IT professionals. He is also interested in the impact of deepfakes on voice and facial biometrics. According to Malinka, the effectiveness of cybersecurity procedures and tools depends heavily on users' understanding of them (or understanding of the consequences of not following them) and on users' willingness to continue to follow them. Malinka himself classifies his research project under the category of usable security and the most general question of "How to design security measures so that they cannot be misused?" When asked where he ranks the title of associate professor in his professional career, he replies: "Associate professorship is a pragmatic personal goal that I value. I have taken over the Security research group, and I need the appropriate academic title to lead the team. Of course, it also gives me the opportunity to continue my research." As for his professional future, Malinka mentions his desire to broaden the scope of AI security. He cites the example of large language models, which pose a number of relatively new security challenges, including prompt injection attacks, data poisoning, and jailbreaking. Taking a longer and bolder view, Malinka mentions the fundamental importance of post-quantum cryptography and the security aspects of hybrid quantum and neural networks as research topics. The number of topics in the field of cyber security will only continue to grow.

Kamil Malinka receives his appointment decree as associate professor from Ladislav Janíček, Rector of Brno University of Technology. | Author: Václav Koníček


Author: Václav Koníček


The second new associate professor at the Faculty of Information Technology is Ondřej Lengál. His habilitation thesis is entitled "Awesome Automata: Algorithms and Applications," clearly referring to his research focus. Lengál works in a field that many believe will see significant developments in the near future. He is an expert in quantum computing, or, as he himself specifies, in verification within quantum computing with the aim of ensuring that these calculations, and by extension programs, do not contain errors. The main goal of his professional efforts is therefore to provide a framework for the automated formal verification of the correctness of quantum programs, which uses automata and logic for the compact representation of complex sets of quantum states with several dimensions of infinity. As Lengál points out, the creation of quantum programs is an extremely demanding discipline, and verification is a desirable part of the development process and a prerequisite for achieving the necessary accuracy and safety of their operation. Quantum programs are inherently probabilistic—at any given moment of their execution, they are in a superposition of states. The various possible states of the program then have different degrees of probability. This makes it easier for programmers to make mistakes. On the other hand, quantum computers are a real promise. Optimistic scenarios predict a quantum boom within the next 10 years. They are expected to be able to effectively solve problems that are very difficult to solve with classical computers. Quantum algorithms promise significantly higher computing speeds, an example of which is factorization in cryptography. After all, breaking the current RSA encryption of web traffic in connection with the advent of quanta is often mentioned as a key cybersecurity challenge in the coming years. And it is also a professional challenge that Lengál wants to devote himself to in the coming years.

Ondřej Lengál is also a new associate professor at FIT VUT | Author: Personal archive of O. Lengál

In the future, the greatest benefit of quantum computers could lie in the simulation of quantum systems in drug development, materials research, and experiments in the fields of physics and chemistry in general. For now, this is just a promise. The state known as quantum supremacy, when we will have a real-world case that has been solved much faster with the help of quantum computing than with classical computers, is currently the most discussed goal. Whether we view the current state of quantum computing with hope or slight concern, it is clear that educating future talent for this field will also be a priority for FIT. Ondřej Lengál also sees the importance of his recent associate professorship in this context: "Quantum experts will be needed. That's why I'm happy about my associate professorship, which will allow me to continue my research and educate new talent."

Congratulations, and we look forward to new successes in research and education at FIT.

Author: Personal archive of O. Lengál

[img]

Page:

Back to top