HomeOps v1
Dishes, bed-making, trash, counters, cabinets, drawers, and everyday object handling.
The real world is the training set. Pick the skill, the environment, and the volume. AuraOne captures it as robot episodes — every clip with signed consent, a safety review, and an export manifest. The data is yours.
The open datasets total about 5,000 hours. Your robot needs more. The open robot-episode datasets total about 5,000 hours against a foundation model’s appetite. We capture the rest, reviewed.
Every clip carries its rights.
Identity-verified consent, a license boundary, and a deletion link travel with each episode — the answer to the EU AI Act provenance clock, which enforces in August 2026.
A second source you can defend.
Independent capture with provenance and safety review attached, so you are not betting your foundation model on a single vendor.
Failure and recovery, not just clean runs.
Drops, blocked views, and wrong-object grasps — with the recovery steps, safety review, and a regression bank so the same failure is caught again.
You keep the data.
It arrives as training-ready episodes with a data card, a manifest, and checksums — not a feed you rent.
A pilot is concrete.
100 to 500 reviewed episodes, $25K to $100K, scoped with you. Submit and a robotics lead replies with a scoped quote and turnaround — not a vague “we’ll be in touch.”
Request one of these as a starting point, or describe a skill of your own in the form above. Failure and recovery capture is a pack, not an afterthought.
Dishes, bed-making, trash, counters, cabinets, drawers, and everyday object handling.
Surface wipe-down, clutter pickup, spills, tools, before/after states, and safe recovery.
Sorting, folding, hampers, drawers, hangers, garment variants, and failed-fold recovery.
Pouring, stirring, opening containers, appliance use, food prep, and safe cleanup.
Picking, packing, scanning, shelving, totes, carts, and exception handling.
Shelf photos, planogram checks, out-of-stock handling, item movement, and display recovery.
Plugging, unplugging, cable routing, port alignment, and rack-style manipulation.
Drops, spills, failed grasps, blocked views, wrong objects, interruptions, and recoveries — labeled so the same failure is caught next time.
Teleop sessions, human-intervention logs, policy-failure tags, and before/after correction capture — the up-the-curve eval data.