Pushing the boundaries of Sim2Real transfer and robotics foundation models.
Our proprietary pipeline collects and curates millions of robot interaction trajectories across diverse environments and task configurations.
Early results from our π0.5 policy rollout trained entirely on our collected synthetic dataset, demonstrating successful sim2real transfer for remote teleoperation tasks.
Advanced domain randomization techniques to ensure robust transfer from simulation to physical hardware across lighting, texture, and dynamics variations.
All demonstration datasets are recorded using our custom teleoperation pipeline. This system captures real-time operator control inputs synchronized with multi-view camera feeds and high-frequency telemetry joint states.
Rollout result (running Autonomously) of the π0.5 VLA (Vision-Language-Action) policy trained on our custom dataset inside simulation.
Talk to the founders about your use case — or see the platform in action with a 30-minute demo.