Home / Data services
Seven services, one pipeline. Teleoperation, demonstration, sensor capture, annotation, synthetic, evaluation, and long-tail capture — built and run on your behalf.
Most foundation-model teams use four or more. We scope per service or as a bundle.
VR, exo, and bilateral leader-follower rigs. Sub-10ms latency, 30Hz joint logging.
Egocentric video, hand pose, gaze tracking, and force gloves for VLA pre-training.
Synchronized RGB-D, lidar, IMU, tactile, and audio. Microsecond alignment.
Action segmentation, affordance masks, language captions, reward signals.
Isaac and MuJoCo scene generation, domain randomization, and sim-to-real bridging.
Held-out test sets, scenario libraries, and scoring infrastructure for model release.
Targeted collection for failure modes identified from your production model logs.
The services chain naturally. You can use one or all seven, and each output feeds the next stage.
Teleop, demonstration, and sensor capture produce the raw trajectories, video, and sensor streams.
Synthetic and sim-to-real expand the captured set with controlled randomization and rare-event coverage.
Annotation adds the structure your model needs: actions, affordances, language, reward.
Eval benchmarks score your trained model. Long-tail capture closes the loop with targeted re-collection.
FAQ
One SOW. One pipeline. One dashboard across every service you choose.