GRADUATE RESEARCH ASSISTANT
🤫
🤫
Research on AI Safety with the CAISA lab at b-it Center, specifically on Emergent Misalignment in the context of Reinforcement Learning and Continual Alignment.
Researched graph reinforcement learning for online combinatorial optimization in the ITF group under the supervision of Dr. Hans-Christian Schmitz and Jonathan Cawalla. Focus on dynamic routing problems such as the Pickup and Delivery Problem.
Worked on a project on online transport and treatment planning for patients in mass casualty incidents. Designed problem-specific heuristic baselines, developed, trained and evaluated graph-based RL policies, and built harnesses that let LLM agents solve the problem directly.
Conducted a literature review on reinforcement learning and metaheuristic approaches for Vehicle Routing Problems for another project, which developed into my bachelor thesis topic. Details in the Education section.
Intelligent Systems Track with focus on Natural Language Processing, AI Safety and Interpretability.
Thesis: Developed a Memetic Algorithm for the Pickup and Delivery Problem with Time Windows, using an agent trained via Reinforcement Learning to intelligently select operators for the local search (Grade: 1.0).