AI Training
Benchmark-grade
AI data,
led end to end.
For frontier AI labs and data partners. I author tasks myself and lead the team that scales them.
Free · 30 minutes · No commitment
2,500+
Tasks delivered
90%
Acceptance rate
4.89/5
Average quality rating
25+
Contributors & reviewers
Hands-on
Tasks I author
50 authored · 45 accepted
Terminal-Bench
Dockerized terminal tasks with reference solutions and deterministic tests.
9 authored · 7 accepted
SWE-bench
Real repository bugs, a fix, and the tests that prove it.
24 authored · 22 accepted
PaperBench
Research papers turned into graded rubric trees.
Quality
Four gates before anything ships
How we start
Pilot first, then scale
01 · Day 1
Share scope
Domain, volume, rubric and timeline.
02 · Week 1
Paid pilot
5–10 vetted experts on a small batch you can grade.
03 · Weeks 2+
Scale
Full volume within weeks, with review built in.
Engagement
Two ways to work together
Individual specialist
From $30 / task
I do the work myself, where quality and accountability matter most.
- Prompt writing & evaluation
- CLI reasoning task design
- Annotation & labeling
- RLHF & red-teaming
- Benchmark creation + test suites
Coordinated delivery
Custom
A vetted team through FynoLabs, with me as your single point of contact.
- Everything in the specialist tier
- Contributor sourcing & management
- Peer review & calibration
- High-volume datasets
- STEM, coding and finance
Next step
Have a dataset to build?
Start with a pilot.
Tell me the domain, task type and volume. We'll scope a paid pilot with 5–10 vetted experts.