I build generalist robot policies. I mostly work on agentic post-training, World Action Models and Vision-Language-Action Models.
What I'm Doing
Stanford
I'm working on post-training of agents/robot policies for bimanual manipulation. I'm interested in a variety of approaches (RLVR, etc.) and I'm always happy to meet new people, so shoot me a message!
microagi
Independent research fellow
What I've Done
Microsoft Spatial AI Lab
I worked on Video Action Models/World Action Models for generalizable robot control.
ETH Robot Learning
We taught a course on sota robot learning. We had 50 SO-101 setups for students that worked in small groups and the best were able to demo their projects in front of Yann LeCun and others. We also had great guest speakers, so look it up:
mimic
Vision Language Action Models (VLAs) and World Action Models (WAMs) for dexterous bimanual manipulation. We did mimic-video, a World Action Model where we propose using a video model as backbone for a robot action policy, because video models have strong temporal/dynamics grounding. This work was published at RSS2026.
NASA JPL
I worked as a robotics intern at NASA JPL on project SWIM, a project on exploring icy moons in our solar system in the search for life. You can spot me in the youtube video on the official NASA JPL channel:
ETH
I completed my bachelor's thesis with ETH Zurich in collaboration with IBM Research, working on model based reinforcement learning for robotic control. The work was published in Nature Machine Intelligence.