The Cyber-Physical Systems research group led by Prof. Matthias Althoff at the Technical University of Munich is offering a PhD/Postdoc position focused on provably safe reinforcement learning for autonomous vessels. The project addresses the growing importance of data-driven approaches in autonomous vehicles, particularly the challenges of training and certifying systems with machine learning components, which are often represented by large neural networks that are difficult to analyze and interpret. To ensure safety without compromising technological advancement, the group proposes integrating machine learning techniques with formal methods. The research will center on safe reinforcement learning for motion planning in autonomous vessels, where the system must not only avoid collisions but also comply with operational rules. Reinforcement learning is chosen for its ability to learn intelligent agent behaviors without labeled data, often surpassing human performance. Instead of verifying neural network correctness directly, the project will develop a safety net that forwards only safe actions to the vessel's actuators, allowing for easy replacement of machine learning components without re-certification. The safety net will verify actions online, simplifying the verification process compared to analyzing large neural networks. Feedback from the safety layer will be used to improve learning rates by excluding unsafe actions during exploration. The project is conducted in collaboration with CargoKite, a company developing ships for autonomous, flexible global container transportation using kite-based propulsion and diesel engines. The new control system will be tested and optimized on a CargoKite ship prototype, with the aim of creating a marketable stand-alone package for various ship types. Applicants will be directly advised by Prof. Matthias Althoff and are expected to have strong scientific and implementation skills, including the ability to lead student teams. The position offers a three-year contract (with possible extension for postdocs) and competitive remuneration according to the German TV-L E13 scale. The Technical University of Munich is committed to diversity and encourages applications from women and disabled candidates. International applicants are welcome. The application deadline is November 15, 2025, with an expected start date between December 2025 and April 2026. To apply, candidates should submit their application via the provided online form, using 'Provably Safe Reinforcement Learning' as the position title and omitting a cover letter. For further information, applicants can refer to the group’s open positions page and previous related publications.