I build the intelligence layer for real-world robots, focusing on software and robot learning systems that enable general-purpose behavior. My work centers on perception-first approaches, ensuring robots can robustly understand and generalize across real-world environments before acting.
What I'm currently working on
I am building an agentic intelligence system called Rogent, deployed on a real-world R2D3 robot. The system uses modular agents – perception, memory, and manipulation – to form grounded representations of the environment. My current focus is on perception and scene understanding as the foundation for action.
Research Interests
- Perception Grounded scene and object representations for real-world robots.
- Manipulation Robust control and interaction under uncertainty.
- Navigation Reasoning and planning in dynamic, unstructured environments.
- Memory State and context across perception and action.
- Reinforcement Learning Learning from interaction and experience in real-world settings.
Contact
For questions or collaboration, please reach out at
[email protected]