About the Role

Build autonomous agents designed to handle complex computer-based work reliably over extended periods. You will own core agent systems, from execution and memory to the evaluations that measure performance against human baselines.

What You'll Do

Build agents that can pursue a goal over multiple weeks with minimal human intervention.

Design memory systems that help agents retain knowledge and improve over time.

Develop evaluation frameworks to determine whether agents outperform people on specific tasks.

Improve reliability and reduce the need for human review of agent runs.

What We're Looking For

Hands-on experience building production agents that people rely on, including a clear understanding of their failure modes.

At least 1 to 2 years of experience at a fast-growing or AI-focused startup.

Experience building agent memory, long-running execution, multi-step reasoning, or systems that maintain goals over time.

A track record building evaluations that inform real decisions, including comparisons against human-level baselines.

Evidence of substantive engineering work, with either published research others build on or software shipped at meaningful scale.

Strong engineering fundamentals; a degree from a highly selective institution or an equivalent credential, such as founding a company, is valued.

Agent architecture experience and backgrounds in competitive math, programming, debate, or games are beneficial.

A collaborative, thoughtful approach to working through difficult problems.

Compensation & Benefits

Base salary is $200,000 to $300,000 annually, with up to $350,000 for exceptional candidates. Meaningful founding equity is included. Visa sponsorship is available.

Location

On-site, five days per week in San Francisco, California, United States.