Founding Engineer, Environments and Data

Headcount: 1

Job Details

• Experience: 0–3 years
• Salary: Not specified
• Equity: 0.25%–2%
• Visa sponsorship: H-1B transfer supported
• Employment type: Full-time
• Work arrangement: In person

Benefits & Perks

• Medical, dental, and vision insurance
• Lunch every day in the office
• 401(k)
• Unlimited PTO
• Paid parental leave
• Unlimited AI spend
• High-ownership environment

About This Role

Our agents do paid work for companies every day, generating production traces and real model failures that public benchmarks miss. You will turn those into products for frontier labs: reinforcement learning (RL) environments, trace datasets, and benchmarks. You’ll own synthetic environments that feel real—for example, a simulated company an agent can work in for weeks, with every email, record, and file generated from scratch—as well as graders designed to resist reward hacking and benchmarks whose rankings match production. You’ll also turn user corrections into benchmark tasks. The founders currently do this work; you’ll take it over. Success means delivering environments that labs buy and train on.

Environment delivery bonuses are paid on top of base compensation and equity: a share of the profit from every environment sold to a lab, up to $500K per year.

Requirements

• Degree from a top-20 school
• Post-training or RL background in research or industry
• Ability to recognize what makes an environment good, including depth, texture, and realistic-feeling assets
• Ability to identify how a model may exploit an environment or grader to produce a technically correct answer that does not meet human intent, and to address it
• Openness to AI tools and active use of them to work more effectively