
AI teams
Labs, robotics companies and ML teams bring us a capability. We bring back the dataset, sourced, verified and licensed.
- Custom datasets to spec
- Expert data annotation and labeling
- Standing orders as you scale
AI teams get custom datasets and expert data annotation. Experts find paid AI work in their field. Owners license their data on their own terms. Every item carries consent and proof of where it came from.
Have data to license? List it on HUMXN
AI teams buy, experts make, owners license. We vet everyone who joins, match the right people to each request, and keep the proof that every item was made and shared with permission.

Labs, robotics companies and ML teams bring us a capability. We bring back the dataset, sourced, verified and licensed.

Physicians, lawyers, engineers, linguists and other specialists find paid AI work that needs the judgement only they can give.

Creators, studios, labs and companies list the data they hold, decide if and how AI may learn from it, and get paid when it does.
The next leap in AI won't come from more scraped data. It will come from better data, made by experts, with permission and proof.

Most data vendors sell what they already have. We start from what you need, then go and find it, and we stay on as your needs change.
You tell us the capability you're training for, where your current data falls short, and the bar it has to clear. We turn that into a written spec: modality, domain, volume, acceptance criteria and licence.
We match the brief to creators who already hold rights to suitable work, and to vetted domain experts who can create, label or grade what doesn't exist yet. You see samples before anything scales.
Every item is fingerprinted, checked for human origin and duplicates, screened for safety and personal data, reviewed by a second expert, and bound to its consent record and signed provenance.
Data arrives licensed, documented and in the format your pipeline expects, with a signed receipt listing every item. Need more next month? Standing orders keep new, matching work flowing.
Verification isn't a badge we add at the end. It's a record built as each item moves through intake, review and delivery, and it travels with the data.
Exact and perceptual fingerprints find copies, crops and re-encodes.
An Ed25519-signed manifest, C2PA where the format allows, and a Bitcoin-anchored timestamp.
Releases, permissions and licence terms bound to the item, not a spreadsheet.
Illustrative record. Every delivered item carries one.
From teleoperation episodes to clinical reasoning, every kind of data goes through the same intake, the same checks and the same proof.
Demonstrations from trained operators, synchronized video, proprioception and force, with task success reviewed by robotics specialists.
Consented recordings with speaker releases, transcripts and expert phonetic labels.
Reasoning traces, rubrics, preference pairs and evaluations written by people who practice the field.
Camera originals with releases, human-made origin checks and dense captions.
Meshes, point clouds and drawings with units and measured-versus-modelled labels.
Egocentric, instructional and cinematic video with frame-level fingerprints and scene annotations.
Contracts, filings and tables with schema, personal-data scans and masking before anything ships.
Training data now gets audited by regulators, litigators and customers. Ours is built to hold up.
Every item has a release or licence from the person who holds the rights.
Nothing is taken from the open web without permission. Ever.
Text, image, audio, video, 3D, documents and robotics data.
Delivered the way your training pipeline already reads data.
Each item is fingerprinted at intake, checked for human origin and duplicates, screened for safety and personal data, reviewed by a qualified expert, and tied to a consent record and a signed provenance manifest with an independent timestamp. You receive a receipt that lists every item by ID and hash.
Quality and trust. Buyers, experts and creators are each approved before they join, and buyers see curated selections rather than an open catalogue. We make the introductions, so every relationship starts with the right people on both sides.
Yes, that's most requests. When the catalogue doesn't cover a brief, we commission it: experts create, label or grade new examples to your spec, and creators produce new work under a licence agreed up front.
Yes. We annotate data you supply or data we source for you: text, documents, images, video, audio and 3D. Specialist tasks go to credential-checked experts in the field, such as radiologists for imaging or attorneys for contracts, and every label records who made it, with a second expert reviewing an agreed share.
Creators keep ownership. You receive a written licence for the uses you need, from AI training to evaluation, with the term and exclusivity you choose. Exclusive licences close the work to other buyers.
Experts are paid for approved work at rates set by the specialty and the task, with payouts made automatically through Stripe once identity and tax details are verified.
Browse open opportunities: paid, flexible gigs that each show their field, pay and time commitment. If nothing fits yet, apply to the expert network and we'll invite you when matching work opens.
Yes. List it with a short description, who holds the rights and whether people in it have consented. A person reviews every listing, and you set the price and licence terms before anything is offered to a buyer.
Yes. We source teleoperation and manipulation demonstrations, egocentric video and sensor recordings from trained operators, with success labels reviewed by robotics specialists.
Send a brief in five minutes. A data lead replies within one business day with questions and a first sourcing plan.