We are researchers building physical AI systems. We train our own policies, build our own data infrastructure, and deploy autonomous systems end-to-end.
Our work spans the full stack: high-fidelity motion capture at scale, synthetic data multiplication, post-training loops, and deployment feedback. We own every layer from raw sensor data to running autonomous operations.
We capture training data using SOTA full-body motion capture pipelines — multi-view video, structured action datasets, and rich metadata ready for policy training. This data feeds directly into our research on manipulation, navigation, and multi-task learning.
phi9.space builds products that accelerate development in physical AI — through data capture, algorithmic innovation, and tools that others can use.
We work openly, in partnerships, and with stakeholders across the physical AI ecosystem.
The goal is transfer — systems that operate reliably in real workflows, with better data, better evaluation, and better transfer from research to deployment.