I study agents, representations, and learning methods that enable machines to re-create and simulate our world. I seek simple, effective methods that generalize across domains.
- Agentic 3D reasoning and generation. Agents that understand and create physically sound, functional, simulatable, and manufacturable 3D assets and scenes.
- Spatial and physical intelligence. Understanding and generating 3D environments and learning to interact with the physical world.
- Foundation model adaptation. Efficient post-training and guidance for foundation and domain-specific models.
I also explore general agent reasoning and generative models.
Related projects
- Agentic 3D reasoning and generation
- BesiegeField, ChatGarment, SGP-Bench
- Spatial and physical intelligence
- PAct, MeshDiffusion, G-Shell, EVA
- Foundation model adaptation
- DrPO, VGG-Flow, PGD, Nabla-GFlowNet, EVA, ASAP, Nabla-R2D3