Egocentric data · VLA annotation · Verified experts

Name any task.
We deliver it
training-ready.

Commissioned first-person manipulation data, captured to your spec by a verified contributor network — VLA-annotated, hand-pose enriched, PII-scrubbed, provenance-hashed, and delivered format-native. Loader-validated on every export.

Pilot bounties start at $300. You write the spec. You pay only for clips that pass your acceptance criteria.

TASK_SPEC — FL-0042COLLECTING
category ····· "hvac_service"
task ········· "condenser_coil_service" reps 15
operator ····· licensed_hvac · cert_verified
capture ······ 1080p60 · hands_visible ≥ 92%
annotation ··· segments · language · success_flags
consensus ···· dual-model gated
pose ········· 21 keypoints/hand · per-frame conf
pii ·········· faces scanned · detections blurred
provenance ··· sha256 → chain, per clip
format ······· LeRobot v3.0 · RLDS · HDF5
validation ··· policy smoke test · loss curve attached
v3.0
LeRobot native · loader-validated
100%
Consent-cleared & hashed
$300
Pilot bounty entry
Per-clip
Pay only for accepted footage
What we collect

The data your model can't scrape, simulate, or collect in-house.

Web video is exhausted. Sim-to-real has a ceiling. In-house collection produces one kitchen, one lighting condition, one pair of hands. We operate the layer in between — real homes, real worksites, real expertise.

Household manipulationRGB · VLA-LABELED

Kitchens, laundry, cleaning, organization, tool use — high-volume egocentric footage across dozens of unique environments per order, segmented and language-labeled.

volume tier
Skilled tradesSCARCE · CREDENTIAL-VERIFIED

HVAC, electrical, plumbing, machining, commercial kitchens — licensed professionals on permissioned worksites, with domain-accurate action vocabulary no crowd platform can produce. The footage nobody else can source.

flagship tier
Evaluation setsEXCLUSIVE · DATED · UNPUBLISHED

Collected-after-date footage held out of every training delivery — enforced in an append-only registry, not promised in a contract. Contamination-free by construction, provable clip by clip.

eval tier
Annotation onlyYOUR FOOTAGE · OUR PIPELINE

Send us your existing recordings. We return temporal action segments, language instructions, success flags, hand-pose tracks, and PII scrubbing — in your training format, with a validation report.

service tier
Pipeline

Every clip runs the same gauntlet.

Raw footage is not training data. Seven automated stages stand between a contributor's phone and your training loop — and the last one proves the dataset actually trains.

01 — INTAKE

Guided capture

Contributors record against your spec with framing guidance and per-task checklists. Technical properties measured on upload — resolution, frame rate, stability, exposure.

ffprobe · capture tiers
02 — QC

Spec scoring

Vision-model scoring against your acceptance criteria — task completion, hand visibility, framing, spec conformance — with confidence bands and human review on the margin.

VLM rubric · confidence-scored
03 — SCRUB

PII removal

Every frame face-scanned; detections blurred with temporal smoothing. Face-clear clips ship untouched with their scan manifest. Originals segregated, never modified.

detect-first · scan manifests
04 — ANNOTATE

VLA action labels

Temporal action segmentation, per-segment language instructions, object references, success flags — generated by a multi-stage vision pipeline and gated by dual-model consensus, with disagreements flagged, never silently shipped.

segments · language · success · consensus
05 — ENRICH

3D hand pose

21 keypoints per hand, per frame, with per-keypoint confidence and honest occlusion handling — gloves and tool-occluded moments flagged, never hallucinated. Commercially-licensed extractor, versioned and regenerable.

keypoints · confidence · coverage stats
06 — PROVENANCE

Hash on-chain

Every accepted clip hashed with its consent record at acceptance. Contributor payouts carry the clip hash in the on-chain memo. The rights chain is independently auditable, clip by clip.

sha256 → chain
07 — DELIVER

Format-native

LeRobot v3.0 (v2.1 on request), RLDS, HDF5/robomimic — episodes, not folders. Full data card: label provenance breakdown, pose coverage, consent references, model manifests.

LeRobot · RLDS · HDF5
08 — VALIDATE

Policy smoke test

Before you ever see it, every delivery loads in the current LeRobot release and trains a reference policy on a held-out split. You receive the loss curve. If it doesn't train, it doesn't ship.

ACT reference run · loss curve attached
Engagement models

Start small. Scale on results.

No minimums, no lock-in, no enterprise sales cycle. Every rung pays only for footage that passes your acceptance criteria.

EVALUATE

Demo slice

Free

A sample of the full stack — annotated, pose-enriched, hashed, format-native, with the smoke-test report attached. Load it in a training run this week and judge the pipeline on your own metrics.

Request the slice →
SPEC

Spec'd slice

Free · one per team

Send your taxonomy and acceptance criteria. We cut a slice against your spec — your categories, your labels, your format — so you're evaluating conformance, not our defaults.

Send a spec →
PILOT

Pilot bounty

From $300 · pay per accepted clip

You write the spec, we fund it to the network. QC-passed, annotated, format-native footage in your hands inside two weeks. Small enough for a corporate card. Real enough to train on.

Scope a pilot →
SCALE

Commissioned collection

From $5K · acceptance-criteria pricing

Hundreds to thousands of hours to your exact spec, on 30–60 day cycles. Quality risk sits with us: you pay for accepted clips only. Retainers and exclusivity windows available.

Scope an order →
Provenance & rights

Data your legal team will actually sign off on.

The bottleneck in physical-AI data isn't pixels — it's rights. We built the consent chain first and the collection network on top of it.

Consent, per clip

Every contributor attests IP assignment, likeness and biometric release, and bystander consent before their first frame — with the consent text version recorded against each clip, not assumed across an account.

indemnification-ready licensing

Scrubbed, then shipped

Every frame face-scanned; detections blurred. Minors auto-rejected. Scan manifests ship with the delivery, so your PII posture is a document you can hand your counsel — not a promise you inherited.

PII pipeline as a legal control

Hashed, verifiable

Each accepted clip is hashed with its consent record at acceptance and the hash rides the contributor's on-chain payout memo. Audit any file in your delivery independently — the payment and the provenance are the same record.

sha256 → chain, per clip
Why spec-driven

Collection farms sell hours. We sell conformance.

There is no shortage of vendors who will point a camera at a kitchen. The question is what arrives in your training loop — and whether you can prove where it came from.

Volume collection networks

  • Generic operator pools — whoever answered the ad
  • No credential layer: nobody on the network can service an HVAC condenser
  • Consent posture inherited from a recruiting form
  • Enrichment sold as "early access" or shipped as raw video
  • "Format-native" as a claim, not a validated artifact
  • Eval contamination handled by promise
  • Payroll runs twice a month — so does operator patience

Firstly Labs

  • Every clip commissioned against a written spec
  • Credential-verified trades: licensed professionals, domain-accurate labels
  • Consent + provenance hashed per clip, at acceptance
  • VLA labels and 3D hand pose on every delivery, disclosed and regenerable
  • Every export loads in current LeRobot — and trains, with the curve attached
  • Eval sets enforced contamination-free in an append-only registry
  • Contributors paid in USDC within the hour — which is why the good ones stay

Your next training run is one email away.

Tell us the task, the volume, and the acceptance criteria. We respond within one business day with a spec draft and a timeline. Or just ask for the demo slice — it's free, and it ships with the loss curve.

data@firstlylabs.com
Request a pilot →