Bridging the Gap Between AI & Physical Intelligence

We research, build, and scale open-source reinforcement learning frameworks to make autonomous robots adaptable to real-world environments.

Our Core Focus Areas

Developing foundational models for robotic manipulation, locomotion, and simulation.

🤖

Sim-to-Real Transfer

Training robust neural networks in ultra-fast physics simulations and safely deploying them onto hardware.

🧠

Imitation Learning

Leveraging human demonstrations and vision-language models to teach robots complex everyday tasks.

🌐

Open Benchmarks

Maintaining reproducible datasets and public environments to measure progress across the scientific community.