Staff + Software Security Engineer, Labs

Staff-plus engineer on Security Labs, an Anthropic team that runs short, high-risk security projects aimed at the protections frontier AI will eventually need, accepting that many will fail. Each project lasts weeks and ends in either a handoff to the team that will run it or a written account of what was learned. Current bets include an isolated high-assurance research cluster, cryptographic proof of which model produced an output, microVM workload isolation, formally verified ML kernel compilation, short-lived regenerating clusters and using Claude to drive security work. The engineer scopes, builds and tests projects with research and infrastructure teams and helps choose the next ones.

Anthropic wants a strong programmer in Python plus Rust, Go or C/C++ with rare depth in hardware security, applied cryptography, kernel or hypervisor internals, formal methods, reverse engineering or cross-domain systems, and a record of self-directed building.