App Data OS · Robotics

The command center for robot learning in the real world. Every run remembered.

Bring in human demonstrations, teleop sessions, failure cases, and expert evals. AuraOne turns them into reviewed, rights-attached datasets with quality scores, data cards, delivery manifests, recall traceability, and eval loops.

Build your robotics dataset See the capture network

Two ways to work with us

Need the data first? Human Data OS captures it: real operators, signed environments, teleop reviewers, failure replay, and expert workflow contributors. This side is where robot data becomes a governed product: reviewed, licensed, packaged, recallable, and ready for eval.

Human Data OS · Robotics

Human Data OS proof inside App Data

Every asset carries its own reason to trust it.

The lifecycle bridge keeps org, operator, task, shift, capture, program, consent, license, quality, upload proof, review state, recall state, and support history on one record.

Task spec

Consent receipt

License scope

Device and sensor metadata

QA score

Reject / rework trail

Annotation and failure labels

Data card

Delivery manifest

Recall / deletion linkage

Eval / regression-bank output

Dataset packs. Delivery surfaces. One product system.

HomeOps v1

Dishes, bed-making, trash, counters, cabinets, drawers, and everyday object handling.

CleaningOps v1

Surface wipe-down, clutter pickup, spills, tools, before/after states, and safe recovery.

LaundryOps v1

Sorting, folding, hampers, drawers, hangers, garment variants, and failed-fold recovery.

FoodOps v1

Pouring, stirring, opening containers, appliance use, food prep, and safe cleanup.

WarehouseOps v1

Picking, packing, scanning, shelving, totes, carts, and exception handling.

RetailOps v1

Shelf photos, planogram checks, out-of-stock handling, item movement, and display recovery.

CableOps v1

Plugging, unplugging, cable routing, port alignment, and rack-style manipulation.

FailureOps v1

Drops, spills, failed grasps, blocked views, wrong objects, interruptions, and recoveries — labeled so the same failure is caught next time.

TeleopOps v1

Teleop sessions, human-intervention logs, policy-failure tags, and before/after correction capture — the up-the-curve eval data.

Four motions. One robotics data platform.

Robot-learning and VLA teams

Build a dataset

Name the skill, the environment, and the format. Get back reviewed, rights-cleared episodes your training stack reads.

Open intake

Approved mobile operators

Run the capture network

One app for operators: find tasks, sign consent, capture to spec, fix rework, and get paid.

Open intake

Location owners and workforce channels

Turn locations into capture sites

Make a home, kitchen, warehouse, hotel, or retail floor a rights-cleared capture site, with owner consent and privacy zones.

Open intake

Expert workers and eval buyers

Evaluate expert work

Score credentialed experts against rubrics and golden outputs, with provenance on every judgment and pay tied to the result.

Open intake

STARTS FROM

A model already suited to the workflow.

STARTER MODEL

OpenVLA

Starts with OpenVLA, LeRobot-style policies, and open robotics models suited to session review and dataset curation.

Operator feedback, consent and license events, review decisions, failure labels, and downstream training results improve the data program around your robot work.

AuraOne can help stand up the first safety, rights, and delivery loop. Your robotics team still owns what becomes training data.

READING · ROBOTICS · LIVE

00·00 INTAKESAFETY REVIEWSIGN 04·18

ROBOTICS STUDIO

Four ways to run the same loop.

Review locally, scale to the cloud, deploy inside your tenant, or hand the second pass to managed reviewers. The dataset stays yours.

OPEN · TIER

№ 01

Robotics Studio Open

A local-first review IDE for teleop, VLA, failure clusters, sensor QA, and reviewed exports.

STARTS FROM

OPENVLA

CLD · TIER

№ 02

Robotics Studio Cloud

Hosted review queues and dataset storage that ship a data card, manifest, and checksums with every export.

STARTS FROM

OPENVLA

ENT · TIER

№ 03

Robotics Studio Enterprise

Self-hosted or VPC robotics data operations with SSO, audit evidence, and signed export attestations.

STARTS FROM

OPENVLA

PRG · TIER

№ 04

AuraOne Robotics Programs

Managed, rights-cleared capture for HomeOps, LaundryOps, WarehouseOps, FailureOps, TeleopOps, and buyer-specific packs.

STARTS FROM

OPENVLA

· TIER

№ 05

Signed Environment Programs

Real homes, kitchens, warehouses, hotels, and retail floors as capture sites, with location-owner consent and privacy zones.

STARTS FROM

OPENVLA

· TIER

№ 06

Expert Workflow Evals

Credentialed experts scored against rubrics and golden outputs, with provenance on every judgment and pay tied to the result.

STARTS FROM

OPENVLA

DELIVERY PROOF · SIX NODES

Not random clips. Robot training data.

The raw clip is only the beginning. AuraOne gives every robotics team the complete record: task, consent, license scope, context, quality check, reviewer decision, data card, recall link, and delivery package.

Review

Human reviewers check the task, context, quality, and release decision before delivery.

Regression memory

Rejected clips can become failure examples that help teams avoid repeating bad behavior.

Safety review

Risky sessions can be held for a safety lead before they enter the training set.

Export formats

RLDS, OpenX, HDF5, BVH, and JSON through the public export surface.

Training handoff

Fine-tuning inputs, weights delivery, and release evidence stay explicit in the program scope.

Delivery package

Raw files, task data, review decisions, signed manifests, and checksums ship together.

HONESTY

Your data stays yours.

HONESTY

Workers consent before each session.

HONESTY

Rejected clips can become failure examples. Not public datasets.

TRAINING DELIVERY

Approved sessions become a delivery package.

Delivered datasets include raw files, consent and license proof, metadata, privacy events, review decisions, failure labels, data cards, manifests, recall linkage, checksums, and supported training packages. No reconstruction needed.

№ 01

Raw videos

№ 02

Task data

№ 03

Device metadata

№ 04

Environment metadata

№ 05

Operator and session metadata

№ 06

Consent receipt

№ 07

License scope

№ 08

Redaction events

№ 09

Review status

№ 10

Accept, reject, and rework reasons

№ 11

Annotation and failure labels

№ 12

Data card

№ 13

Recall and deletion linkage

№ 14

Eval and regression-bank output

№ 15

Accepted clip list

№ 16

Export manifest

№ 17

Checksums

№ 18

Training format package

↳ NATIVE FORMATS

MP4/MOVJSONLCSVParquetData cardManifestRLDSOpenXHDF5BVHJSON

READING · DELIVERY · LIVE

BUILT FOR

Teams turning raw real-world data into robot skill.

From autonomous vehicle labs to warehouse automation teams, the same evaluation surface scales across company types and fleet sizes.

↳ TEAM TYPE

Humanoid robotics teams

Teams collecting human movement, household tasks, workplace motion, and failure examples for physical AI programs.

↳ TEAM TYPE

Autonomy research labs

Research groups that need reviewed raw video, movement data, teleop sessions, and export packages before training changes.

↳ TEAM TYPE

Embodied AI programs

Teams building vision-language-action policies from real people doing real tasks — not lab reconstructions.

↳ TEAM TYPE

Teleoperation research groups

Groups that need structured teleoperation sessions and reviewer decisions for downstream training.

CATEGORY MAP · WHERE WE SIT

What this is and is not.

A short read on the landscape. Anonymized by category, not company — because the point is the work, not the logo.

Fast-capture worker networks

Fast raw clips. AuraOne adds task design, consent, rights scope, quality checks, reviewer decisions, data cards, and delivery manifests before training.

Generic gig-labor marketplaces

Massive worker supply, weak robotics context. AuraOne is built for signed robot training data, reviewer-graded sessions, rework trails, and payout lineage.

Open public datasets

Useful starting points. AuraOne helps you build your own proprietary dataset on top of the public baseline.

Internal operations

Spreadsheets, folders, and Slack threads do not scale a robotics data program. The work needs one lifecycle record, reviewer queue, delivery manifest, and recall path.

TRUST & SECURITY · WHAT TEAMS ASK

Serious data, simply handled.

QUESTION

Who owns the data?

Your data stays yours. We do not resell customer clips. The accepted clip list, manifests, and weights you tune all belong to your program.

QUESTION

Do workers consent?

Workers consent before each session and see the task, license summary, privacy requirements, and review terms they are being asked to follow. The receipt is part of the record.

QUESTION

What happens to rejected clips?

Rejected clips are not wasted. They can become failure examples for your robotics team — not public datasets.

QUESTION

How are exports delivered?

Exports are packaged for robotics teams with raw files, task data, consent receipts, license scope, review decisions, data cards, signed manifests, recall links, and checksums attached.

QUESTION

Can you work with our existing teleop stack?

Yes. Teleop sessions can connect to the customer program, provider setup, and physical environment, with the capture plan scoped before the work starts.