Robot Intelligence

We advance robot intelligence through foundation models and world models that enable robots to perceive, reason, and act in complex physical environments. Vision-Language-Action (VLA) models allow robots to ground language understanding in real-world manipulation and locomotion. Our world models learn physics-aware representations of environments, supporting sim-to-real transfer and long-horizon planning. We apply these approaches across manipulation, humanoid control, autonomous vehicles and racing, and multi-agent systems — combining multimodal sensing (vision, tactile, proprioception) with deep reinforcement learning and computer vision.