Humanoid Robotics · Physical AI · Machine Learning
Graduate student in robotics and machine learning at TUM, interested in the foundations of machine learning for physical AI.
I'm a graduate student at the Technical University of Munich in robotics and machine learning, currently focused on the foundations behind physical AI and humanoid robots — vision-language-action models and world models.
I hold a Bachelor in Engineering Science from TUM, during which I had the privilege of several research and exchange stays abroad — at the Politecnico di Milano, the University of Zurich, and CentraleSupélec in Paris. Today I'm a robotics engineer with RoboTUM and help run partnerships at the European Student Robotics Association (ESRA).
I like work that spans the full arc — from the research idea to a system that actually runs on hardware. Past projects have ranged from on-board ML for a CubeSat to voice-driven motion planning on a Unitree G1 humanoid.
I'm always glad to talk about robotics, physical AI, and ML research. The fastest way to reach me is email.
Focus areas
Toolbox
Languages
P/01
A feed-forward pipeline that turns short indoor phone videos into geometrically coherent 3D scenes with open-vocabulary semantic labels — no COLMAP or per-scene optimization. Built for a humanoid perception & spatial-AI challenge, it supports natural-language 3D queries in milliseconds.
P/02
Team work at RoboTUM on the Unitree G1 humanoid robot, where I led the voice pipeline and motion planning — connecting spoken commands to planned, executable motion on the robot. Developed collaboratively with the team.
P/03
A non-invasive wearable that classifies sleep position from two accelerometers using machine learning — 100% accuracy across the four main position groups (supine, prone, left, right). End-to-end hardware: PSoC microcontroller, I²C sensors, and Bluetooth streaming.