Evals Researcher, Rewards and Environments

Evals Researcher (Rewards and Environments)

Location San Francisco, CA Base $150,000–$350,000 plus equity Type Full-time, hybrid

$150K–$350K

base salary, plus equity

Hybrid

3–4 days a week in the

office

6-day week

the standard work week

Full-time

contract type

About the role

An early-stage AI lab in San Francisco is hiring an Evals Researcher to design the rewards, graders and

environments used to train and measure frontier models. You'll build evaluation systems that run on real traffic

and decide what "good" means for the model.

What you'll do

• Build graders, model-as-judge systems and RL environments, and run them on production traffic.

• Design reward signals and checkers that hold up when models try to game them.

• Work with the research and engineering team on RL data, training runs and evaluation pipelines.

• Decide what to measure, and say clearly what you think should be worked on next.

You have

• Hands-on RL data or RL research experience, built yourself and not adjacent to your work.

• An evaluation system you built that ran on production traffic or sits behind a published result.

• Real opinions on reward design: what makes a checker trustworthy, and how models game it.

• Strong systems engineering skills and the ability to pick up any stack, with an engineering title in your

current role.

• At least one year of full-time experience (one to five years is typical).

Details

• Six-day work week; hybrid in San Francisco, with three or four days in the office.

• The company