Security Engineer / Research Engineer, RL Security

Secures the reinforcement learning systems that teach Claude to reason, code and act, working inside Anthropic's RL Engineering team. A frontier RL run spans training, sampling and environment execution across a large accelerator fleet, and this hands-on engineer designs and runs the security controls for it across orchestration, environments, training and inference.

The work includes reviewing designs and code for security flaws, building shared libraries and observability, and checking with researchers that changes keep training correct and fast. Anthropic hires both security engineers drawn to ML systems and ML engineers drawn to security; it wants Python plus Rust, C++ or Go, production distributed or security systems experience, and Linux, container and cloud knowledge.