In detail
- XDOF: raised $70M from Thrive, Spark, a16z, Lux, WndrCo; ~60 employees; working with ~20 customers
- XDOF focus: collect, pipeline and annotate high‑fidelity robotic interaction data lacking in public sources
- ENPIRE (Nvidia/CMU/UC Berkeley): eight dual‑arm YAM stations; agents reach up to 99% success on tricky tasks
- ENPIRE approach: agents automate resets, success checks and reward functions; agents read papers, form hypotheses and edit training code
Why it matters
Robots need high‑quality, domain‑specific interaction data and automated training loops to scale; businesses planning physical automation should budget for data collection and consider new tooling that reduces human overhead.
For you Assess your data needs for any robotics pilots and explore partnerships with specialist data providers; evaluate agent‑driven training methods to reduce manual intervention in deployments.