Our goal is to enable Embodied AI and robot learning inspired by the human learning process. The current work focuses on foundation models for building efficient scene representations and simulation environments as well as robot learning in these environments. We are also interested in novel perception and sensing modalities. Overall, our research spans basic problems in machine learning, computer vision, and robotics. Current interests involve the following topics:

• Robotics: Large-scale Models for Robotic Perception and robust Policy Learning.
• Computer Vision: Efficient & Editable Neural 3D Reconstruction, Generative Scene Representations, Neuromorphic Vision.
• Machine Learning: Geometric Deep Learning, Causal Representation Learning, Reinforcement Learning, Imitation Learning.