Security Engineer

Objective

Make agent authority explicit before execution and make containment decisive when behavior crosses that authority.

What you’ll own

• Agent identity and resource-scoped authorization
• Threat models for agent-to-tool, agent-to-agent, and agent-to-data paths
• Security review of evidence and approval posture before wider rollout
• Containment, root cause, and follow-through for agent incidents

Evidence of success

• Each privileged action is attributable to an agent identity and granted scope
• A security reviewer can reconstruct the decision and evidence behind an action
• Containment closes every known execution path during an incident

Constraints you should know

• Human approval cannot repair missing machine-enforced scope
• Audit evidence must come from the execution path, not a later reconstruction
• Security controls must remain usable under incident pressure

Working style

• Work in person with the founding team in San Francisco
• Read implementation code and own the operational response
• Turn threat models into shipped controls and regression tests

What we look for

• Shipped identity, PAM, or capability-based authorization in production
• Written threat models other teams used to ship changes
• Drove or co-led an incident response that closed a real exposure
• Read code in two of Go, Rust, Python, and TypeScript