General Analysis
Research Engineer, Evaluations
San Francisco, CA · Full-time · Posted Oct 29, 2026
About the job
Build evaluation and security infrastructure for frontier AI models, including agent frameworks, adversarial evaluation environments, and tools for analyzing results.
What you'll do
- Develop agent frameworks and scaffolding for delegated, multi-step tasks; design evaluation environments, including challenges measuring evasion of detection.
- Own the evaluation pipeline end to end, from API design to analysis and visualization of agent trajectories.
- Collaborate with security experts to ground evaluations in real-world practice, and share findings through publications and customer presentations.
What they ask for
- Strong production programming skills.
- Strong problem-solving and analytical skills, including extracting signal from messy data.
Nice to have
- Experience building agentic systems, evaluation harnesses, or environment frameworks, or eagerness to learn.
- Security experience, or eagerness to build it.
AI focus
Develop agent systems and evaluation infrastructure to assess and strengthen frontier AI models on cyberdefense tasks, including adversarial testing and detecting evasion.
Model-written summary · the original posting is authoritative