Stefan Pranger

Stefan Pranger is a PhD candidate in the Trusted AI Group at TU Graz, supervised by Prof. Bettina Könighofer. His research focuses on the formal analysis and enforcement of safety properties for AI systems, with a particular emphasis on reinforcement learning under uncertainty. He develops methods that combine probabilistic model checking, runtime monitoring and enforcement, specifically shield synthesis to provide safety guarantees for learning-based agents. His recent work also addresses testing of deep reinforcement learning and explainability of synthesized shields.

Stefan is the main developer of Tempest & tempestpy. Additionally, he is actively developing a tool for autonomous experiments on scanning tunneling microscopes (STMs).

Research Interests

• Safe reinforcement learning and shield synthesis
• Probabilistic model checking and quantitative verification
• Enforcement, testing, and explainability for AI systems