Research

I’m a doctoral researcher in Hector Geffner’s group (Chair of Machine Learning and Reasoning, i6) at RWTH Aachen. The group’s broad aim is combining learning with model-based reasoning and planning. My corner of that is robots and long-horizon tasks — keeping what’s learned tightly grounded in symbolic structure rather than left implicit.

A few connected questions I work on:

• Geometry-aware guidance for task-and-motion planning — helping a planner generalise across tasks while staying aware of what is physically feasible.
• From symbolic plans to learned control — how abstract plans connect to learned low-level skills so they actually execute.
• Representations between symbol and geometry — learning representations that move between the abstract and the concrete.

Software

I build and maintain TAMPanda, a MuJoCo-based task-and-motion-planning library (inverse kinematics, RRT* motion planning, PDDL task planning, with Gymnasium integration for RL), and contribute to the ROS 2 CLIPS-Executive for rule-based robot coordination.