Research · Remote · 1 week
Eval & Benchmark Curation Intern
Help sharpen how we test Kestrel-Nano by writing evaluation prompts and scoring rubrics for coding, tool-use, and instruction-following.
What you'll actually do
- Write 15–25 test prompts in an assigned category (e.g. "reads a stack trace and picks the right fix").
- Draft a short pass/fail rubric for each prompt.
- Score a batch of real model outputs we provide against your rubric and flag disagreements.
Good fit if
You like precise writing, enjoy finding edge cases or holes in an answer, and can read a short code snippet — you don't need to write production code.