Jim Fan is a Director of AI and Distinguished Research Scientist at NVIDIA, where he co-founded and co-leads the GEAR group (Generalist Embodied Agent Research) with Yuke Zhu. He is best known for his work on embodied AI, robot foundation models, and humanoid robotics, including helping lead Project GR00T. Before NVIDIA he completed a PhD at Stanford under Fei-Fei Li and was OpenAI’s first intern. His current focus is foundation models for general-purpose robots in virtual and physical environments.
Recent
-
2026-07-07 SimFoundry paper (NVIDIA GEAR + Fei-Fei Li’s team): Real2Sim system that converts a single real-world video into an interactive robot sim environment. Policies trained inside it transfer zero-shot to real robots with 17-40% task success improvements. Mean Pearson correlation 0.911 between sim and real-world performance across 7 manipulation tasks. arXiv 2606.28276 · NVIDIA
-
2026-06-16 Unveiled ENPIRE: 8 Codex agents given a robot fleet, GPU budget, and token budget to run physical experiments autonomously — resetting scenes, reading papers online, patching control stacks. Solved zip-tie tying, fine-pin organization, and GPU installation without human help. Claims first “AutoResearch in the physical world.” X