Skip to content

Jacksonville, Florida

danni@curryrecruit.com
Curry Recruit

All jobs

Hiring employer: Confidential

Evals Researcher, Rewards and Environments

San Francisco, California · Hybrid · Full time

$150,000 to $350,000 base, depending on experience and seniority, plus meaningful equity

Apply Now (opens in a new tab)

The hiring employer is confidential and is not named on this page.

This evals research role at an AI research lab builds the evaluation systems that decide which data is worth training on and whether models and agents are actually getting better. The work covers reward design, automatic scoring at scale, and catching drift as traffic and task shapes change. You will rank training value before compute is spent, and design scoring cheap enough to run constantly.

The schedule is six days of work a week, hybrid in San Francisco, typically three or four days in the office.

Requirements

  • Has built reinforcement-learning data or done reinforcement-learning research, hands-on, not adjacent
  • Has built an evaluation system (graders, model-as-judge, environments, or training pipelines) that ran on production traffic or behind a published result
  • Real, substantive opinions on reward design: what makes a checker trustworthy, and how models learn to game it
  • Comfortable owning ambiguous evaluation problems end to end
  • Able to work six days a week, in person or hybrid in San Francisco
  • Must already hold the right to work; no new sponsorship (the bar is higher for anyone on OPT or needing a petition)