Offensive Cybersecurity Engineers: Build RL Environments for AI Evaluation

Role overview

Build and strengthen offensive cybersecurity environments used to evaluate advanced AI systems. The work combines hands-on security engineering with benchmark design, developing scenarios that can support rigorous, ongoing assessment of model capabilities and risks.

Responsibilities

- Design and implement reinforcement learning environments for offensive security scenarios. - Harden existing benchmarks against sophisticated interactions by AI systems. - Define evaluation criteria that make security assessments rigorous and repeatable. - Work iteratively with collaborators to update scenarios as vulnerabilities and evaluation needs evolve.

Requirements

- Professional experience in offensive cybersecurity, penetration testing, or security research. - Strong practical skills building or hardening technical environments. - Ability to work on a remote, hourly project over an extended period. - Interest in AI safety and evaluating model capabilities.

Nice to have

- Experience developing security benchmarks or secure test environments. - Familiarity with reinforcement learning or AI evaluation frameworks.