Member of Technical Staff

Responsibilities

At Frist, you will:

• Build benchmarks spanning an entire use case, e.g., 300 tasks inside the domain of accounting, law, healthcare
• Improve automated application generation pipelines
• Design tasks that are realistic, long-horizon, and hard for frontier models
• Write rubrics and reward functions that grade multi-step computer workflows, where "did it work?" is rarely a single check.
• Create synthetic data pipelines that turn a handful of human demonstrations into thousands of training-grade trajectories.
• Instrument everything: environment cost, throughput, failure modes, where agents get stuck and why.

We’re looking for

• Strong communication, engineering skills
• Ability to learn quickly and build fast
• Curiosity about AI infrastructure, post-training open-source models, task design, and rubric building.
• Prior ML experience is welcome, not required

Nice to have

• Post-training experience with LLMs: RL with verifiable rewards, reward modeling, or building evals.
• Computer-use or browser agents, or browser automation at scale (Playwright, Puppeteer, CDP).
• Code-generation pipelines that produce working software with little human help.
• Sandboxing, virtualization, Kubernetes or large-scale job orchestration.
• You've worked at an RL-environment or AI data company, or founded your own company.

What we offer

• Compensation: Competitive salary plus meaningful early stage equity.
• Location: In person in New York City, 5 days a week.
• Benefits: Full healthcare coverage, gym, plus meals, coffee and more.
• Scope: Founding-team ownership over systems frontier AI will train on.