For AI labs & data vendors
Expert data you can audit, item by item.
- RLHF, preference data and expert rewrites
- Held-out evaluations and red-teaming
- RL environments and verifiable tasks
- Specialists staffed onto your own projects
Leaderboard
8 AI models vs. a 1908 Urdu text: the best got 1 word in 11 wrong
Expertise AI can't scrape. Verified before it trains.
The vetted expert workforce for frontier AI labs and AI-data vendors. Physicians, lawyers, scientists, engineers and native speakers of 100+ languages, each identity- and credential-checked before they write, rank or red-team a single item of your data.
For AI teams
Request a pilot
For experts
Apply as an expert
- 551
- open roles
- 142
- languages hiring
- 33
- fields hiring
- Clinical medicine
- Pharmacology & drug safety
- Law & regulation
- Physics & chemistry
- Aerospace & space
- Software & mathematics
- Defence & security
- Finance & accounting
- Agronomy
- Education
- Art & design
- Tibb & traditional medicine
- العربية Arabic
- हिन्दी Hindi
- বাংলা Bengali
- اردو Urdu
- Kiswahili
- Bahasa Indonesia
- Hausa
- Tagalog
- Vietnamese
- English
Two ways in
One network of verified experts. Two sides of the same trust.
AI teams get data from people whose identity and credentials are on record. Experts get fair, paid work that uses what they actually know.
For experts
Paid, remote AI work in your own field.
- Rate agreed per project, before you start
- Paid in US dollars by bank transfer, Payoneer or Wise
- A written contract and consent over how your work is used
- No fee to apply, ever
Proof, published
We gave AI models a 1908 Urdu text. The best still got 1 word in 11 wrong.
A vetted native expert typed the truth blind, every disagreement was checked against the scan, and every word was scored. On our public leaderboard the top model, Claude Fable 5.1, reproduced 91.08% of words exactly and still made a mistake on 111 of 161 lines. Many errors carry no warning flag, so nothing in the output tells a reviewer to look twice.
- 91.08%
- best strict word accuracy (Claude Fable 5.1)
- 111/161
- lines with a mistake, even for the best model
- 2,378
- words and marks typed blind by a vetted native expert
- 8
- models ranked so far, more added as they finish
The same method runs in any field: truth written blind by a vetted expert, adjudicated against the source, scored item by item.
What we deliver
Data that makes models measurably better at the hard parts.
Managed pods of verified experts, scoped to your capability gap and quality bar.
Expert RLHF & preference data
Rankings, rewrites and written rationales from credentialed professionals, not generalists guessing.
Evaluations & red-teaming
Gradable test sets and adversarial probes in medicine, law, science, engineering, finance, defence and 100+ languages.
RL environments & verifiable tasks
Tasks with checkable answers and reward signals, designed by the people who do the work.
Reasoning in 100+ languages
Native speakers with subject expertise, from major world languages to low-resource ones, not machine translation of English prompts.
Provenance-licensed datasets
Consented, documented and exclusive-ready, with a chain of custody for every item. Built for a legal landscape where provenance decides what can be used.
How verification works
Five checks before anyone touches your data.
Every expert passes all five before client work, and the result of each check is kept on record, so you can see who produced your data and why they were qualified to.
-
Identity
Government-ID verification with consent, a liveness check and duplicate-account detection.
On record: ID match
-
Credentials
Degrees, licenses and registrations checked with the issuing body: medical boards, bar associations, engineering institutions, universities.
On record: registry check
-
Domain assessment
A role-specific test, written and graded by senior practitioners in the field.
On record: score vs. pass mark
-
Paid calibration
Trial tasks scored against gold answers before any client work begins.
On record: agreement with gold
-
Continuous quality
Ongoing agreement scoring and spot audits. Anyone below the bar comes off the project.
On record: per-batch audit
Provenance by default
- Consent recorded for every contributor
- Chain of custody for every item
- Community-consent protocol for cultural material, following the CARE principles
- Exportable audit trail with each delivery
Security roadmap
- SOC 2 Planned
- ISO/IEC 27001 Planned
- ISO/IEC 42001 Planned
Designed for least-privilege access, encrypted storage and per-project isolation from day one. We only claim a certification once an auditor has issued it.
Fair work
- Written contracts for every contributor
- Rate agreed per project
- Payouts on a published schedule
- No fee to apply, ever
For AI teams
From capability gap to delivered data in four steps.
-
Step 01
Scope
Tell us the capability gap, the domain and your quality bar. We design the task and the rubric with you.
-
Step 02
Match
We assemble a pod of verified experts whose credentials fit the task, not whoever is online.
-
Step 03
Produce
Dual review, gold tasks and calibration on every batch, with agreement scores you can see.
-
Step 04
Deliver
Data in your format, with its provenance ledger attached.
Domains
Every field a frontier model is asked about. Experts who can check it.
We recruit and vet experts across the professions, the sciences and the world's languages: wherever a model's answer needs a qualified human to write it, grade it or break it.
Medicine, pharma & Tibb
Clinical reasoning, diagnostics, pharmacology and drug safety, plus evidence review of Tibb-e-Unani and other traditional medicine.
Law & policy
Statutory and case-law reasoning, contracts, IP and patents, and compliance, across common-law, civil-law and religious systems.
Sciences
Physics, chemistry, biology and mathematics: proofs, derivations and reasoning rubrics written by researchers.
Engineering & software
Civil, electrical, mechanical and software engineering; code review, systems and technical protocols.
Aerospace & space
Aeronautics, propulsion, flight systems, avionics and space operations.
Finance & business
Banking, capital markets, accounting, tax, insurance, strategy and operations, including Islamic finance.
Defence & security
Strategic analysis, cybersecurity, OSINT and adversarial red-teaming of model behavior.
Education & academia
Teachers, professors and assessment writers for curricula, exam items and step-by-step explanations.
Art & design
Visual art, design, music, film and architecture, judged by working practitioners.
Languages
100+ languages: native-speaker reasoning, translation QA, transcription and dialect coverage, from major world languages to Urdu and other regional languages.
Agriculture & climate
Agronomy, crop and soil science, irrigation, veterinary science and climate adaptation.
Heritage & humanities
History, philosophy, religious and literary traditions, manuscripts and archives, documented and consented.
Vettedlance research
Measuring what frontier models don't know yet.
The Urdu study is published. Three more expert-built benchmarks are in development, and each will publish after independent expert review, with methodology and agreement scores.
Statutory reasoning across jurisdictions
Statute and case-law questions graded against verifiable citations, jurisdiction by jurisdiction.
- Format
- Questions with verifiable citations
- Written by
- Practising lawyers
- Scoring
- Per item, citation-checked
Low-resource language reasoning
Instruction-following and reasoning in languages with little presence in training data, written by native speakers.
- Format
- Native-written prompts and reference answers
- Written by
- Native speakers with subject expertise
- Scoring
- Blind, adjudicated
Clinical & pharmacology safety
Clinical reasoning and traditional-medicine claims graded against evidence and flagged for safety.
- Format
- Clinical and drug-safety cases
- Written by
- Physicians and pharmacists
- Scoring
- Evidence-graded, safety-flagged
Why it matters beyond Urdu: frontier models score up to 24.3 points lower when the same questions are asked in a low-resource language (MMLU-ProX, 2025), and 88% of the world's languages were found to be left behind by language technology (ACL, 2020).
Why Vettedlance
Built for trust, not volume.
| Criterion | Generic crowd platforms | Vettedlance |
|---|---|---|
| Who does the work | Whoever is online | Practitioners matched to the task by field and language |
| Identity | Self-reported | Government-ID verification with consent |
| Credentials | CV claims | Checked with the issuing body |
| Quality control | Spot checks after delivery | Gold tasks, calibration and agreement scores on every batch |
| Provenance | Rarely documented | A record for every item: who, when, which guideline, how reviewed |
| Coverage | Generalists, English-first | Recruited by domain, in 100+ languages |
| Contributor terms | Opaque, variable | Written contract, rate agreed before work starts |
Open roles
551 ways to put your expertise to work. Paid in dollars.
Physicians, lawyers, scientists, engineers, analysts, artists and native speakers, across 33 fields and 142 languages. Apply once and get matched to paid, remote AI projects.
Physician (MBBS/FCPS) for Clinical AI Evaluation
Paid in USD
Medicine & health
Expert · Talent pool
Lawyer (Advocate) for Legal Reasoning Evaluation
Paid in USD
Law & policy
Expert · Talent pool
Software Engineer for AI Code Evaluation
Paid in USD
Engineering & software
Expert · Talent pool
Mathematician (MS/MPhil/PhD)
Paid in USD
Sciences
Expert · Talent pool
Unani Medicine Physician (BEMS/MD Unani)
Paid in USD
Traditional & Eastern medicine
Expert · Talent pool
Mufti / Fatwa Research Scholar (Iftaa)
Paid in USD
Religion, Sufism & folklore
Expert · Talent pool
Aerospace Engineer for AI Evaluation
Paid in USD
Aerospace & space
New opportunity
Clinical Pharmacologist for RLHF
Paid in USD
Pharma & life sciences
New opportunity
Defence Analyst for AI Red-Teaming
Paid in USD
Defence & security
New opportunity
Urdu Transcriptionist, Historical Texts
Hiring nowPaid in USD
South Asian languages · اردو
Transcription · Open now
Swahili Language Expert for AI Training & Evaluation
Paid in USD
African languages · Kiswahili
Expert · Talent pool
Arabic (Modern Standard) Language Expert for AI Training & Evaluation
Paid in USD
Middle Eastern & Central Asian languages · العربية
Expert · Talent pool
For experts
Your expertise, paid in dollars. On time.
If you hold a professional license, an advanced degree, or native fluency in a language models handle badly, your knowledge is exactly what AI systems are missing.
- Rate agreed per project, before you start
- Paid in US dollars by bank transfer, Payoneer or Wise
- A written contract and consent over how your work is used
- Flexible, remote work that fits around your practice
- No fee to apply, ever
- 01
Apply
About ten minutes. Tell us your field, credentials and languages.
- 02
Verify
Confirm your identity and credentials with your consent.
- 03
Assess
A short test written by senior people in your field.
- 04
Calibrate
Paid trial tasks so you learn the quality bar.
- 05
Work
Join projects that match your expertise.
FAQ
Questions, answered plainly.
Are you already working with AI labs?
We're onboarding our founding expert cohort and building our first datasets and benchmarks. We don't list clients or partners we haven't signed, and we never will.
Can we start with a pilot?
Yes. Most teams start with a fixed-price pilot scoped to one capability gap, then scale once the quality is proven on their own evaluation.
How do you verify experts?
Five checks: government-ID verification with consent, credential checks with the issuing body, a domain assessment written by senior practitioners, paid calibration tasks, and continuous quality scoring. Nobody touches client work until all five are complete.
Do you work with AI-data vendors as well as labs?
Yes. We can staff verified specialists onto a vendor's own projects, or deliver finished data and evaluations, with the same verification and provenance records.
How do you handle copyrighted or cultural material?
We only use material we have the right to use, and we document where every item came from. For cultural and community knowledge, we follow the CARE principles for Indigenous data governance and record consent before anything is collected.
How are experts paid?
In US dollars, at a rate agreed before a project starts, by bank transfer, Payoneer or Wise, on a published schedule. Every contributor signs a written contract.
Which domains and languages do you cover?
We staff by demand rather than a fixed list: medicine and pharma, law, the sciences, engineering including aerospace, software, finance, defence, education and the arts, plus English, the major world languages and the low-resource languages where native-speaking experts are hardest to find. We add a language once we can verify enough native speakers with subject expertise to meet our quality bar.
Is my data secure?
We design for least-privilege access, encrypted storage and per-project isolation from day one. SOC 2, ISO/IEC 27001 and ISO/IEC 42001 are on our roadmap, and we only claim a certification after an auditor issues it.
Get started
Put verified expertise behind your model.
Start with a scoped pilot, or join the expert network. Either way, verification comes first.
For AI labs & data vendors
Scope a pilot in your domain, languages and quality bar. We reply within two business days.
For experts
Paid, remote AI work in your field. Free to apply, with a written contract for every project.
Sources for figures on this page
- Vettedlance, blind human evaluation of a frontier model on a 1908 Urdu Nastaliq text (method, results and records)
- MMLU-ProX multilingual benchmark, 29 languages, arXiv 2503.10497 (2025)
- Joshi et al., The State and Fate of Linguistic Diversity and Inclusion in the NLP World, ACL (2020)
- CARE Principles for Indigenous Data Governance, GIDA