Co-founded Skild AI after a decade+ in robot learning research at CMU; focuses on technical development and deployment of general-purpose robot intelligence.
Building a universal 'omni-brain' for robotics analogous to LLMs for language — pretraining on diverse video and simulation data, then post-training on real robot data to create a data flywheel across verticals from factories to homes.
Deep technical partnership across simulation (Isaac Sim, Newton), generative video models (Cosmos), and on-device edge compute — critical because robots need immediate reaction without server round-trips.