01FIELD
Perception
See the world as a space for action.
Transforming rich visual and spatial signals into grounded representations that agents can use.
- Vision
- 3D scenes
- Multimodal sensing
Researching intelligence in motion
Where perception
becomes action.
I explore embodied intelligence at the boundary of vision, reasoning, and control—building agents that learn through interaction with the physical world.
VISION-LANGUAGE-ACTIONROBOT LEARNINGWORLD MODELSSPATIAL REASONINGVISION-LANGUAGE-ACTIONROBOT LEARNINGWORLD MODELSSPATIAL REASONING
01 / RESEARCH FIELD
Three questions, one embodied system.
Intelligence becomes meaningful when it can close the loop between understanding a situation and changing it.
02FIELD
Reasoning
Connect intent to physical possibility.
Studying how embodied agents plan, remember, and reason through uncertain, changing environments.
- World models
- Planning
- VLM agents
03FIELD
Learning
Improve through interaction.
Building systems that turn experience, feedback, and demonstrations into robust behavior.
- Robot learning
- Adaptation
- Control
02 / THE CLOSED LOOP
Intelligence is a cycle, not a snapshot.
The most interesting agents do more than recognize what is in front of them. They form a hypothesis, act on it, observe what changed, and get better the next time around.
- 01SenseCollect multimodal signals
- 02InferBuild a world model
- 03ActExecute with precision
- 04AdaptLearn from the outcome
03 / OPEN CHANNEL