Materials · case/board/corvid-research-scientist.md

# Corvid Labs: Research Scientist, Agent Evaluation

San Francisco or remote (US) · Full time

Corvid Labs builds evaluation environments for AI agents that operate software: agents that book travel, file expense reports, and work across browser tabs. Labs and enterprises use our environments to decide whether an agent is ready to deploy.

## What you will do

- Design and build evaluation environments and tasks for computer-use agents.
- Develop graders that score long, multi-step agent trajectories, and determine when those graders can be trusted.
- Run studies that compare agent performance with human performance on the same tasks.
- Write up methods and findings for customers and for publication.

## What we are looking for

- PhD or equivalent research experience in a quantitative field.
- Strong Python.
- Experience designing experiments and measurement, and analyzing the results.
- Experience with language models in a research setting.
- Preferred: experience building reinforcement-learning environments or training pipelines.

## How to apply

Apply through our applicant-tracking portal. Upload a résumé (PDF). A cover letter is optional. Applications are screened by our research team, who read every letter.