01SFT
Supervised demonstrations
Prompt and response pairs, and the reasoning that got there, written by people who do the work for a living.
- Prompt and response pairs
- Chain-of-thought reasoning traces
- Multi-turn dialogues
- Per-item author provenance
3,000+ vetted experts, 10,000+ more on call
Elite experts for training frontier models
Doctors, lawyers, engineers and analysts who still practise, producing the demonstrations, rubrics, environments and evals that frontier models learn from.
- SFT
- RL rubrics
- Agent environments
- Computer use
- Evals
The record, since May 2026
- Tasks delivered
- 5,000+
- Accuracy on delivered work
- 99.52%
- Experts on the bench
- 3,000+
- More on call
- 10,000+
The problem
Today’s models are good at producing something that looks like an answer. Real work is mostly judgement: what to check first, which number looks wrong, when a rule does not apply. That judgement lives inside practitioners, not on the internet. We get it out of them and into your data.
Why it is hard
An operations problem disguised as a research problem.
Finding two hundred practising radiologists is not the hard part. Keeping quality flat while the bench turns over, for nine months, against a rubric that keeps moving, is the hard part, and it is not the thing a data lab raised money to do.
That is the part we run.
The catalogue
Five things, and nothing else.
If a request does not land in one of these we say so, rather than learning on your budget.
01 / 05Swipe
Two tracks
Behind your contract, or under ours.
Most of our work is invisible by design. A data lab wins the contract on its platform and research; we supply the bench behind it.
The bench
10 domains, staffed by people who still practise.
A tax expert writing transfer-pricing traces files returns for a living. The moment someone stops practising, they stop being the thing the model needs.
- 01
Medicine and clinical
- Physicians
- Nurses
- Radiologists
- Clinical coders
- 02
Law
- Litigation
- Corporate and M&A
- Intellectual property
- Compliance
- 03
Accounting and tax
- Statutory audit
- Direct and indirect tax
- Transfer pricing
- Controllership
- 04
Finance
- Equity research
- Credit analysis
- Trading operations
- FP&A
- 05
Software and data
- Backend
- Infrastructure and SRE
- Analytics engineering
- QA automation
- 06
Engineering
- Civil and structural
- Mechanical
- Electrical
- Process and quality
- 07
Supply chain and trade
- Freight and logistics
- Customs brokerage
- Import and export documentation
- Procurement
- 08
Insurance
- Underwriting
- Claims adjudication
- Actuarial support
- 09
Pharma and regulatory
- Clinical trial operations
- Regulatory submissions
- Pharmacovigilance
- 10
Business operations
- Recruiting
- Customer operations
- Sales operations
- Bookkeeping
Process
From first call to a batch shipping weekly.
The pilot is paid, priced at the volume rate, and the output is yours whether or not it continues.
- 01
Scope
One call with whoever owns the data. We read your spec, your rubric and a sample batch, then tell you which parts we can staff this month and which we cannot. The second answer is the useful one.
- 02
Pilot
A small paid batch at the full quality bar, in whichever system you chose, through your QA gate. You keep the output whether or not it continues, and the price is the price it would be at volume.
- 03
Staff
Named experts, not a queue. You see the bench before it starts: who they are, what they have practised, how long, and who leads the squad. You can reject anyone without giving a reason.
- 04
Run
Fixed weekly throughput, one point of contact, and daily visibility into what shipped. Sampling and double-scoring run inside every batch rather than after it, so a drift shows up in days rather than at the end of the month.
- 05
Audit
Every batch ships with its agreement statistics, its reject rate, and the items we threw away and why. If a number moves in the wrong direction, you hear it from us before you find it.
Tell us the domain you are stuck staffing.
One call, and you get a straight answer about whether we can take it. If we cannot, we will say so on the call rather than three batches in.
Prefer to read first? See who runs Xocket Data Labs or every way to reach us.