Research Scientist — Evaluation

About this work

For researchers who want to shape how model capability is measured in their own field, rather than have it measured by generalists.

What you’ll do

• Design evaluation problems at the frontier of your field
• Assess model reasoning for subtle domain errors
• Identify misconceptions a non-specialist reviewer would miss
• Advise on what a meaningful benchmark in your field looks like

What we’re looking for

• A PhD or equivalent research experience in any discipline
• Active familiarity with current work in your field
• No AI background required

Time commitment

5–15 hours per week, projects typically run 4–10 weeks

Pay and how you’re paid

$85–$150 per hour, paid every two weeks by bank transfer

How to apply

Apply once and we’ll consider you for this role and anything else that fits. The full process is set out on How it works.