Research
My work focuses on computational algorithmics and complexity, especially as they relate to symbolic AI. In areas like AI planning and system verification, many decision and optimization problems can be represented using logic. While these problems are theoretically challenging, modern solvers often handle real-world cases with surprising efficiency. I am interested in understanding the underlying structures that make this possible, and in developing new algorithms that leverage these insights—with strong theoretical guarantees and practical implementations. Since many applications require not just a solution but an explanation or a count of possible answers, counting plays a central role in my research. My main interests include:
- Facets and Navigating Solution Spaces: Solution spaces of configuration problems or in planning instances can quickly become large. I am interested in understanding and navigating such spaces by quantitative methods. I research general knowledge representation and reasoning techniques that apply for any solution spaces, their computational complexity, and establishing alternative methods to navigate and reason on the search space (e.g., facets).
- Model Counting and Quantitative Symbolic AI: Model counting is central in probabilistic reasoning, understanding solution spaces, and computational physics (weighted versions). I am interested in algorithms, solvers, proof systems for exact and model counting, including massively parallel and certified counting. I am a founder and co-organizer of the Model Counting Competition. Furthermore, I work on determining the complexity of counting and probabilistic reasoning in various symbolic AI formalisms, among them answer-set programming (ASP), formal argumentation, epistemic logic programming, description logic, and planning.
- Parameterized Complexity and Algorithmics: Structural parameters such as treewidth yield algorithms whose running time is exponential only in the parameter, i.e., they work well on instances with hidden structure. I am interested in symbolic reasoning such as ASP, QBF, QCSP and solving algorithms, their limitations by conditional lower bounds focussing, and relative complexity (precise reductions that hold regardless of complexity theoretical assumptions such as the ETH).
- Competitions, Evaluation, and Reproducibility: Empirical evaluation is essential for advancing combinatorial solving, though it requires thoughtful design. Events like the annual Model Counting Competition (started in 2020) and PACE 2019, which I co-chaired, provide valuable opportunities to benchmark solvers. Other initiatives include the Pseudo-Boolean Competition, the MaxSAT Evaluation 2026, ARQ-COMP 2027 (focused on automated reasoning for quantum systems), and the 2025 Dagstuhl Seminar on Competitions and Empirical Evaluations in Automated Reasoning. As the field evolves, questions about how to evaluate solvers reliably are becoming more important—especially given challenges like concurrency on modern hardware. This has led to more nuanced approaches to reproducibility. Moving forward, I am interested in establishing better standards and protocols for algorithm evaluation, along with developing statistical tools that provide deeper insights than traditional worst-case analysis. For a complete list of events, please see the Competitions page.
- Applications: Knowledge representation and reasoning techniques play a key role in many real-world applications, especially when it’s important to explain results or explore a wide range of possibilities. For instance, we explore their usage in the design of future space missions. In aviation, I collaborate with Airbus to apply counting techniques that clarify and navigate complex planning problems. My experience also includes data analysis for health care in the SAHRA research project. Looking ahead, I plan to work on verifying cryptographic systems and network security software, as well as advancing automated reasoning for quantum circuits and programs, which is the focus of the ARQ-COMP competition.
Projects
| Period | Project |
|---|---|
| 2025–2029 | Model-Based Attention in AI Planning Co-PI. Wallenberg AI, Autonomous Systems and Software Program (WASP). PI: Daniel Gnad. |
| 2021–2024 | NAVAS: Navigation im Lösungsraum von Answer Sets Collaborator. Federal Ministry of Education and Research (BMBF). PI: Sarah Gaggl. |
| 2019–2022 | HYPAR: Hybrid Parameterized Problem Solving in Practice (P32830) Collaborator. Austrian Science Fund (FWF). PI: Stefan Woltran. |
| 2019–2022 | Nichtklassische Logiken: Parametrisierte Komplexität und Enumeration Collaborator. German Research Foundation (DFG). PI: Arne Meier. |
| 2018–2019 | Paralleles SAT-Solving (HO 1294/11-1) German Research Foundation (DFG). PI: Steffen Hölldobler. |
| 2015–2018 | SAHRA: Smart Analysis Health Research Access PI of a sub-project. Federal Ministry for Economic Affairs and Energy (BMWi). |
| 2014–2018 | Decodyn: Treating Hard Problems with Decomposition and Dynamic Programming (Y-698) Project assistant. Austrian Science Fund (FWF). PI: Stefan Woltran. |
| 2014–2018 | Complex Reason: The Parameterized Complexity of Reasoning Problems (ERC 239962) Project assistant. European Research Council (ERC). PI: Stefan Szeider. |