THE NETWORK

Earn through work that trains the world's most ambitious AI.

Quartz Labs sources from universities, recording studios, language NGOs, robotics teleoperator networks, credentialed domain experts, content libraries, and dataset aggregators. If you have a network we don't, we want to pay you for it.

Join the network → How partners earn ↓

Network

Who we work with.

Sample partner categories across the modalities we cover. The list isn't exhaustive — if your organization isn't here, apply anyway. We're additive, not narrow.

Partner type	What we need	Modality
Universities & research institutions	Transcription, dialect annotation, recording at university studios, domain evaluation, RLHF data collection.	Voice · RLHF · Eval
Recording studios & language NGOs	In-country native-speaker recording sessions, ethically collected with signed releases.	Voice
Robotics teleoperator networks	Teleop demonstration sessions on standard manipulation, locomotion, and household setups.	Robotics
Credentialed domain experts	Medical, legal, financial, scientific evaluation of frontier AI model outputs. NDA, audit trail.	Expert eval
Content libraries & archives	Licensable podcast catalogs, video archives, broadcast material, ethnographic recordings.	Multimodal · Voice
Dataset aggregators	Pre-collected datasets with clear chain of custody and commercial-cleared rights.	Cross-modality
Others	If you have a network we don't and it helps train frontier AI, we want to hear from you.	Any

Compensation

How partners earn.

Indicative ranges per partner type. Final rate depends on language scarcity, domain complexity, data rights model, and operational tempo.

Partner type	Rate range	Payment cadence	Engagement structure
Universities & research	$40–200 / hr varies by domain	Standing per-semester contracts.	Department-level contract. Multi-project scope.
Studios & language NGOs	$25–60 / hr per session	Per-deliverable, daily payouts.	Per-session or per-batch. Renewable.
Robotics teleoperator networks	$20–45 / hr	Per-session, weekly.	Per-cohort contract. Multi-session standard.
Credentialed domain experts	$60–200 / hr varies by domain	Hourly tracked, weekly.	Per-engagement. Recurring pools for ongoing eval.
Content libraries & archives	Per-asset license case-by-case	Negotiated per asset.	Bulk licensing. Renewable.
Dataset aggregators	Per-dataset case-by-case	Negotiated.	Case-by-case.

Rates are indicative ranges. Final rate depends on language scarcity, domain complexity, data rights model, and operational tempo. Rates can be higher than the range for rare languages or specialized domain experts.

How partnership works.

From application to first payout in two weeks. No unpaid trials. No exclusivity. No lock-in.

Apply.

Five-minute application form. Tell us what network you cover, what modality, your current capacity, and your preferred engagement structure.

Match.

We pair you with a project that fits your network and capacity. Paid pilot first. Pilot scope agreed up front.

Onboard.

We share spec, quality bar, consent template, and payment setup. Production starts within two weeks of pilot acceptance.

Earn.

Daily payouts for studios. Weekly for experts. Standing contracts for university partners. Per-asset or negotiated for archives and aggregators.

Apply now →

Apply

Join the network.

Book a 20-minute partner intro. Tell us what network you cover, your modality, and your capacity — we respond with a paid pilot proposal within 48 hours if there's a fit.

Frequently asked.

If your question isn't covered here, book a partner intro and ask us directly.

How are rates determined?

Rates depend on language scarcity, domain complexity, rights model, and operational tempo. The ranges shown above are indicative. We quote per engagement after reviewing your network and the buyer's spec.

How does payment work?

Bank transfer or international wire. Daily payouts for studios. Weekly for experts. Standing semester contracts for university partners. Per-asset for archives. We do not pay in cryptocurrency.

Do you require exclusivity?

No. Work with other vendors. We're additive.

How is contributor data protected?

Contributor identities are anonymized in every shipped deliverable. Demographic metadata is aggregated, never personally identifying. All sessions are governed by signed consent with documented opt-out at any point. Raw recordings and identifying metadata are stored encrypted and access-restricted, separate from the cleared deliverables that ship to buyers.

Who owns the IP?

Buyer (the AI lab) owns the trained dataset. You retain ownership of your underlying recordings or evaluations unless otherwise negotiated. Every row in every deliverable has a signed consent log tied to its contributor.

What categories of work are you actively expanding?

Underrepresented-language voice across multiple regions, robotics teleoperator cohorts across Asia, credentialed evaluator networks across high-stakes domains, and multimodal pipelines on request. We expand to new languages, regions, and modalities based on buyer demand and partner availability — if you cover something we don't, that's exactly the reason to apply.

What if I refer a partner?

Referral fee on first paid engagement. Book a partner intro and mention the referral.

What if my organization isn't in the partner categories above?

Apply anyway. The categories are samples, not a closed list. We're additive, not narrow.