Small AI labs are building their own models, and they need data from people who know a subject cold. You work through real tasks in your field. We turn that into the training data their models learn from. Paid, remote, on your own time.
The work
Small labs train their models with reinforcement learning, and that signal has to come from real experts. We build a custom environment for each lab (a coding repo, a trading model, a legal review), you work through it the way you would on the job, and the model learns from how you handle it. Your expertise becomes their training data, and you do all of it on rustBench, our experts platform.
Real coding tasks and agent traces — the kind of bugs you'd hit at work.
Modeling and analysis, plus the judgment calls that trip up general models.
Build evals, write rubrics, and review training data and model pipelines.
Walk through physics, chemistry, and math reasoning one step at a time.
Contracts, case law, and policy work, where a wrong answer costs something.
Clinical reasoning, diagnosis, and medical literature, where accuracy is everything.
How it works
A short application. Show us real work — a GitHub, a paper, a portfolio, whatever proves you know your field.
A small paid task in your area so we can see how you think. No unpaid take-homes, ever.
We match you to a lab and you pick up projects on rustBench, our experts platform. Work through their environments, get paid by the hour. Remote, on your schedule.
Paid $20–100+/hr, remote, about three minutes to apply.
Apply to rustBench