AI Center Seminar - AI Fundamentals series - Prof. Emir Demirović

The talk is organized by the EPFL AI Center as part of the AI fundamentals seminar series.

Hosting professor: Prof. Clément Pit-Claudel

Title
Transparent AI by Design: Search Algorithms for Supervised Learning, Control Policies, and Combinatorial Certification

Abstract
AI methods—such as those used in supervised learning, controller synthesis, and combinatorial optimisation—have demonstrated immense value across many domains. However, their practical adoption is hindered by reliability concerns, particularly when these systems are designed as black boxes. Two key challenges arise for black-box AI: (1) lack of performance guarantees—when AI fails, it is unclear whether the task is infeasible or the underlying algorithm is simply inadequate; and (2) lack of confidence—results may be difficult to interpret or trust. While post-hoc interpretability techniques offer partial remedies, we advocate for a different paradigm: building AI systems that are transparent by design. Rather than explaining opaque decisions after the fact, we synthesise outputs that are intrinsically understandable and verifiable. This shifts the focus from doubting AI to questioning whether we are solving the right problem. We apply this approach across three distinct domains: supervised learning, controller synthesis, and infeasibility certification for combinatorial optimisation problems. Although these tasks involve exponentially large search spaces, recent advances demonstrate that designing for transparency is increasingly practical—often without sacrificing performance—making it a compelling alternative to opaque AI systems. A variant of this talk will be presented at ECAI'25 as an invited talk in the Frontiers in AI series.

Bio
Dr. Emir Demirović is an assistant professor of computer science at TU Delft (Netherlands), where he leads the Constraint Solving ("ConSol") research group and directs the Explainable AI in Transportation ("XAIT") lab. He has been recognised with the Early Career Researcher Award 2025 from the Association for Constraint Programming and is an ELLIS Scholar. His research focuses on exploiting structural properties of NP-hard problems to design algorithms that are both theoretically complete and efficient in practice, with a particular emphasis on constraint programming and dynamic programming techniques. His techniques have advanced state-of-the-art solvers in MaxSAT and constraint programming (achieving high rankings in competitions such as the MaxSAT Evaluation and the MiniZinc Challenge), his optimal decision tree methods (machine learning) are among the fastest, and his recent approach to certifying outputs of constraint programming solvers on large-scale problems has been highlighted by Donald Knuth in The Art of Computer Programming (Vol. 4, Fasc. 7) as a promising advancement.