Keynote Speakers
ISD 2026 is proud to feature distinguished Keynote Speakers who are leading experts in information systems development. Their insights will provide valuable perspectives on the latest trends, challenges, and innovations shaping the field.
Title: Semantic Data Interoperability in Practice: Lessons from Public Administration and Beyond
Speaker: Jakub Klímek
Affiliation: Charles University, Prague, Czechia
About the speaker: Jakub Klímek is an Associate Professor at the Faculty of Mathematics and Physics, Charles University, since 2024, where he conducts research and teaching in semantic technologies, open data, and data interoperability, with a particular focus on linked open data and semantic web technologies. Since 2013, he has been active in both academic and public-sector initiatives related to semantic interoperability and information systems development, and, since 2022, he has been the CEO of a Charles University spin-off company CUIT, focused on consultancy projects in the same area.
Alongside his academic work, he has long served as an expert on open and linked data within the Office of the Chief eGovernment Architect, currently within the Digital and Information Agency of the Czech Republic. In this role, he has contributed to the development of the Czech National Open Data Catalogue and to the definition of open formal standards for interoperable data publication in public administration.
Jakub Klímek is the author or co-author of dozens of scientific publications in international journals and conference proceedings and has contributed to major European semantic data specifications developed within the European Commission’s SEMIC initiative, including DCAT-AP, GeoDCAT-AP, DCAT-AP HVD, and StatDCAT-AP. He has recently joined the W3C Dataset Exchange Working Group as Invited Expert. He has served on the program committees of international conferences and workshops, and acts as a reviewer for scientific journals.
Abstract: Digital transformation in the public sector increasingly depends on the ability to exchange, interpret, and reuse data across organizational and technological boundaries. While the promise of interoperable and machine-actionable public data has been discussed for decades, practical implementation remains challenging. This keynote reflects on the Czech experience with semantic interoperability and Linked Open Data (LOD), examining obstacles encountered and approaches taken.
Beyond technical data exchange lies a closely related challenge: achieving a shared understanding of what data actually means, i.e. semantic interoperability. Although Semantic Web technologies are well established, their application in cross-border and cross-domain data exchange continues to raise new organizational and methodological challenges.
The second part of the keynote will focus on semantic data specifications in the context of the European Union and the Common European Data Spaces, addressing how such specifications can be reused consistently across national and domain-specific contexts. This will be illustrated through the example of DCAT-AP, the Interoperable Europe solution for metadata exchange.
The keynote will conclude with reflections on how the development and adoption of semantic interoperability solutions may be supported by AI, and conversely, how semantic technologies and structured knowledge may provide essential foundations for AI systems.
Title: The Great Convergence: Recommendation, Search, and Conversation in Real-World AI Systems
Speaker: Pavel Kordík
Affiliation: Czech Technical University in Prague & Recombee, Czechia
About the speaker: Pavel Kordík is an Associate Professor at the Faculty of Information Technology, Czech Technical University in Prague, where he also served as Vice Dean for Development and Industrial Collaboration between 2016 and 2024. He received his Ph.D. in Artificial Intelligence from CTU in 2007, with a focus on meta-learning and neural network optimization. He is active member of the Data Science Laboratory and the Recombee Research Lab. He is co-founder of Recombee, a company delivering AI-powered recommender systems and personalized search systems as a service to companies with hundreds of millions of active users worldwide. He has co-founded companies and non-profits bridging academia and industry, including prg.ai, experts.ai, edumatch.ai, and the children’s AI-literacy non-profit aidetem.cz. He is the Principal Investigator of multiple EU research projects (FOCAL, PoliruralPlus, Capttict) and recently served as the General Co-Chair of the 19th ACM Conference on Recommender Systems (RecSys 2025) in Prague. Author of more than 90 research publications, his work spans recommender systems and personalization, machine learning and neural networks, meta-learning and AutoML, and AI allignment. Throughout his career, he has been dedicated to fostering synergistic relationships between academic research and industry applications, championing openness, knowledge sharing, and diverse international teams.
Abstract: For most of their history, recommendation, search, and conversational interfaces have been built as separate systems, with distinct architectures, evaluation methods, and engineering teams. This separation is rapidly disappearing. Embeddings, foundation-model backbones, and large language models are pushing these three paradigms toward a single underlying stack, in which a query, a click, a scroll, and a natural-language request are increasingly different entry points into the same retrieval-and-ranking machinery. This talk examines what such convergence means in practice for the applications we build, and the new challenges it brings, both theoretical and practical. On the theoretical side, unifying interaction-based and semantic signals raises hard questions about evaluation, cold-start behavior, bias and alignment. On the practical side, deploying these systems in the real world forces an uncomfortable trade-off: they must be not only accurate, but also fast and economically viable at the scale of hundreds of millions of users — a constraint that reshapes architectural decisions far more than benchmark leaderboards suggest. Drawing on experience from both academic research and the production systems behind Recombee, the talk explores how to scale converged recommendation–search–conversation systems while keeping latency and cost under control, and how to ensure user satisfaction and AI alignment across very different domains, from news and media, through e-commerce, to education. The central argument is that the next generation of information systems will be defined less by raw model capability and more by how thoughtfully we engineer the interaction between accuracy, affordability, and alignment.
Title: Towards controlled and sustainable AI in Model-Driven Development: Guard Rails based on Normalized Systems Theory
Speaker: Herwig Mannaert
Affiliation: University of Antwerp, Belgium
About the speaker: Herwig Mannaert graduated as an electronics engineer in 1988 at the K.U.Leuven. He obtained his PhD at the K.U.Leuven in 1993, specializing in the design of object recognition algorithms for image
interpretation. As a pilot application for these algorithms, he led the development of a real-time video system for license plate recognition that became operational in 1992. From 1992 to 1997, he
taught a course on signal processing, random signal analysis, and digital filter design, leading to the publication of his first textbook in 1998.
In 1993, he became an associate professor at the Management Information Systems Department of the University of Antwerp, where he has been teaching courses at the interface between information and telecommunication technology since then. These courses led to the publication of an overview textbook on electronics in 2005. From 1998 to 2000, he was a lead-developer of the secure communication software for the Society for Worldwide Interbank Financial Telecommunications.
In 2000, he co-founded Cast4All, a company that licenses scalable and transactional software management systems for distributed applications on heterogeneous and often poorly connected
networks. Applications included the distribution of digital movies via satellite to cinemas and cable head-ends, the central management of facility equipment in large datacenters, and the
management of power solutions on optical fiber networks and along high-speed railroads.
Currently, he is a Professor at the University of Antwerp, where he previously served as Vice-Dean, as well as Chairman of the Management Information Systems Department, , and a Director of Cast4All. His main research focus is on the Normalized Systems Theory (NST), a theory on the behavior of modular structures under change. To realize the theory in practice, he co-founded in 2011 NSX bv, a spin-off company of the University of Antwerp focusing on the development of software systems based on NST. He is also a guest lecturer at the Czech Technical University and an IARIA fellow.
Abstract: Normalized Systems theory uses concepts such as stability from systems theory to study the evolvability of modular structures. Its focus is on eliminating so-called combinatorial effects from modular structures, which are ripple effects whose number of impacts is related to the size of the modular structure, which makes them highly undesirable from the point of view of controlled and sustainable evolution of for example, software architectures. To build Normalized Systems in practice, fine-grained skeletons of modular structures are generated and continuously re-generated over time by code generators implementing the principles of Normalized Systems Theory and Model-Driven Development. This approach is very well suited as a complement to the use of AI and LLMs for code generation, as the skeletons can be easily extended with micro-doses of AI-generated code, while at the same time
providing guard rails for the use of AI. These guard rails add traceability and control not only to the code generation process (which are currently seen as major limitations of LLM-based code generation), but also to the continuous re-generation of AI-generated software architectures over time.