Reinforcement Learning Engineer (Cybersecurity)

Overview

Build reinforcement learning environments and verification systems that enable AI agents to discover, exploit, and remediate software vulnerabilities using deterministic, verifiable rewards.

What you'll do

• Design and deploy high-throughput RLVR environments for evaluating LLM action sequences.
• Build test harnesses, sandboxes, and verification engines for vulnerability research.
• Integrate automated analysis platforms and vulnerability feeds into RL environment pipelines.
• Architect isolated, reproducible environments for parallel exploitation and patching trajectories.
• Implement telemetry, verification algorithms, and trajectory logging to detect reward hacking.
• Build instrumentation and debugging tools for memory, process, and network monitoring.
• Benchmark AI models across fuzzing, exploit generation, and patch validation tasks.

What you'll need

• Understanding of RL training workflows for modern LLM systems, including RLVR.
• Proficiency in Python and C programming.
• Understanding of software vulnerabilities, binary exploitation, fuzzing, or program analysis.
• Experience with DevOps pipelines and reproducible builds using Docker, BuildKit, or Nix.
• Comfort working with Linux systems and low-level debugging.
• Experience building or working with benchmark environments, CTFs, or execution sandboxes.

Nice to have

• Rust experience.
• Experience designing reward functions, ground-truth verifiers, or automated grading engines.
• Background with sanitizers, compiler instrumentation, or automated exploit generation tools.
• Experience with cybersecurity benchmarks, CTFs, or open-source AI evaluation frameworks.

Details

• Remote, work-from-home 100% of the time.
• US national base salary range: $176,400–$242,550 yearly.