Data Scientist, AI Model Evaluation (Remote Contract)

Remote, contract, hourly. We are lining up experienced data scientists and quantitative analysts to stress-test how well large AI models reason about statistics, experiments, ML methodology and quantitative problems.
In practice the work looks like this:
• You read a model's answer to a data-science question and decide whether the reasoning actually holds up.
• You write hard, realistic problems (with worked reference answers) that a strong analyst could solve but a sloppy one would get wrong.
• You flag methodological mistakes: bad experimental design, misuse of a test, leakage, over-claiming, and similar.
• You write feedback that a modelling team can act on.
Who this suits
People who have done this for real: at least a year in a serious data, research or finance team within the last several years, and comfortable explaining why an analysis is wrong in plain, precise writing. A degree from a well-regarded university helps. You need to be located in an English-speaking country for this engagement.
Practical details
- Fully remote, you set your own hours
- Independent-contractor arrangement, part-time
- Paid hourly; rates for this engagement sit in the mid-$200s per hour
Applying takes a couple of minutes through OpenTrain, the marketplace where AI-evaluation specialists keep one profile and pick up projects like this one. If your background fits, you will hear back quickly.

Work arrangement
Yes