Theory of Software Systems
Computer Science
1 Career
- Since 2009 Professor, Institute of Science and Technology Austria (ISTA)
- 2009 – 2022 President, Institute of Science and Technology Austria (ISTA)
- 2004 – 2009 Professor, Computer and Communication Sciences, EPFL, Switzerland
- 1999 – 2000 Director, Max-Planck Institute für Informatik, Saarbrücken
- 1998 – 2004 Professor, EECS, University of California, Berkeley, USA
- 1997 – 1998 Associate Professor, EECS, University of California, Berkeley, USA
- 1996 – 1997 Assistant Professor, EECS, University of California, Berkeley, USA
- 1992 – 1995 Assistant Professor, Computer Science, Cornell University
- 1991 Postdoctoral Scientist, Université Joseph Fourier Grenoble, France
- 1991 PhD, Computer Science, Stanford University, USA
- 1987 Dipl.-Ing., Computer Science, Johannes Kepler University Linz
- 1986 M.S., Computer and Information Sciences, University of Delaware, USA
1.1 Selected Distinctions
- 2025 Disruptive Idea Award, Conference on Neurosymbolic Systems (NeuS)
- 2024 Member, Royal Society (UK)
- 2021 Influential Paper Award, Autonomous Agents and Multi-Agent Systems (AAMAS)
- 2020 Member, National Academy of Sciences (USA)
- 2020 Member, American Academy of Arts and Sciences
1.2 Selected Grants
- 2022 – 2026 Austrian Science Fund SFB Grant F8502, Interface Theory for Security and Privacy. Awarded to RG: € 284,040.
- 2021 – 2026 ERC Advanced Grant 101020093, VAMOS: Vigilant Algorithmic Monitoring of Software. Awarded to RG: € 2,451,280.
1.3 Tech Transfer (Patents, Spin-Offs)
N/A
1.4 Selected Publications
- M. Lechner, R. M. Hasani, A. Amini, T. A. Henzinger, D. Rus, and R. Grosu, “Neural circuit policies enabling auditable autonomy,” Nature Machine Intelligence, vol. 2, no. 10, pp. 642–652, [Online]. Available: https://doi.org/10.1038/s42256-020-00237-3.
- D. Zikelic, M. Lechner, A. Verma, K. Chatterjee, and T. A. Henzinger, “Compositional policy learning in stochastic control systems with formal guarantees,” in Advances in Neural Information Processing Systems 36, 2023. [Online]. Available: https://doi.org/10.48550/arXiv.2312.01456.
- U. Boker, T. A. Henzinger, N. Mazzocchi, and N. E. Sara¸c, “Safety and liveness of quantitative properties and automata,” Logical Methods in Computer Science, vol. 21, no. 2, 2025. [Online]. Available: https://doi.org/10.46298/lmcs-21(2:2)2025.
- F. Cano, T. A. Henzinger, B. Könighofer, K. Kueffner, and K. Mallik, “Fairness shields: Safeguarding against biased decision makers,” in 39th AAAI Conference on Artificial Intelligence, AAAI Press, 2025, pp. 15 659–15 668. [Online]. Available: https://doi.org/10.1609/aaai.v39i15.33719.
- C. W. Barrett, T. A. Henzinger, and S. A. Seshia, “Certificates in AI: learn but verify,” Communications of the ACM, vol. 69, no. 1, pp. 66–75, 2026. [Online]. Available: https://doi.org/10.1145/3737447.
2 Group Members
- Mahyar Karimi, PhD Student since 2023
- Pavol Kebis, PhD Student since 2024
- Fabian Kresse, PhD Student since 2024 (jointly supervised with Christoph Lampert)
- Konstantin Kueffner, PhD Student since 2020
- Stefanie Muroya, PhD Student since 2022
- Leonard Brice, Postdoc since 2025
- Filip Cano, Postdoc since 2025
- Ana Oliveira da Costa, Postdoc since 2023
3 Research
We work on designing automata- and logic-based formalisms for modeling and specifying software systems, and on developing algorithmic tools for the analysis of these models and specifications. We focus on concurrent systems, whose complexity arises from the interaction of multiple processes or agents, and on quantitative models and specifications, which take into account how likely it is that a model satisfies a specification, how robustly a model satisfies a specification, or whether a model satisfies a specification in real time. Our methods can be used in security applications, where attackers try to break specifications, in the verification of cyber-physical systems, which have software interact with the physical world, in responsible AI, to ensure the safety and fairness of AI-based decision makers, and in quantum computing, to analyze the hardware-dependent accuracy of quantum algorithms. Since most software is beyond the reach of formal design verification, much of our work develops methods for identifying and correcting software errors at runtime.
3.1 Current Scientific Projects/Directions
- Graph games for modeling autonomous agents
- Quantitative automata
- Monitoring logical safety and statistical fairness
- Verifying quantum systems
- Verifying, certifying, and shielding AI
- Hyperproperties for ensuring security and robustness
3.2 Research in Focus
Quantum hardware is inherently fragile and noisy. We found that the accuracy of traditional quantum error correction algorithms can be improved depending on the hardware. Given different hardware specifications, such as those published by IBM, we automatically synthesized hardware-optimal algorithms for parity correction, qubit resetting, and Greenberger-Horne-Zeilinger state preparation, which are common subroutines in quantum programs. Using stochastic techniques from computer science, such as the algorithmic exploration of Partially Observable Markov Decision Processes (POMDPs), our method presents the first computational tool for computing the exact accuracy (i.e., probability of success) of a quantum algorithm executed on specific hardware. We are the first to synthesize quantum algorithms that are optimally tuned to the hardware; perhaps surprisingly, they often turn out to be different from the textbook algorithms for solving the same problems. We also showed that improvements can be gained with respect to the IBM Qiskit transpiler as we computed the hardware-optimal qubit mappings for tasks such as GHZ state preparation. This work was published together with Stefanie Muroya and Krishnendu Chatterjee in PNAS 122(12) e2419273122 (2025).