Research
Embodied intelligence, end to end
We combine theory with physical systems. Open tools. Published findings. Machines that learn by doing.
Embodied Cognition
Developing AI architectures where learning emerges through physical interaction, not just data ingestion.
Sim-to-Real Transfer
Closing the reality gap with domain randomization, adaptive simulation, and online fine-tuning.
Multi-Agent Coordination
Enabling fleets of heterogeneous robots to collaborate on complex, dynamic tasks.
Human-Robot Interaction
Designing intuitive, safe interfaces for humans and machines to work side by side.
Edge Intelligence
Running sophisticated AI models on resource-constrained embedded hardware.
Robust Perception
Building vision and sensing systems that perform reliably in unstructured environments.
Philosophy
The hardest problems in AI cannot be solved in simulation alone. Our method pairs rigorous theory with hardware that fails in public.
We share tools like Synapse SDK and publish openly. Progress in robotics is too important to hoard.
