Agentic AI Security Engineer

What you'll do

• Develop and maintain hardened execution environments for agent code using kernel and VM-level isolation.
• Implement network egress controls and sandbox security on OpenShift platforms and HPC clusters.
• Build guardrails to enforce policies on agent inputs, outputs, and tool calls.
• Deploy judge models and human approval gates to manage agentic behavior.
• Create independent safeguard services for routing and monitoring agent sessions.
• Establish systems for attributable audit logging and real-time monitoring of agent activities.
• Perform post-incident analysis and conduct adversarial testing on the safety stack.
• Make foundational architecture decisions for the security infrastructure of AI agents.

What we're looking for

• Minimum 5 years of experience in security, systems/SRE, or ML systems engineering.
• Hands-on experience operating LLM agents and managing their failures.
• Deep expertise in sandbox or container security with the ability to apply it to agentic domains.
• Proficiency in Linux and Kubernetes/OpenShift isolation techniques.
• Working knowledge of agent architectures including tool-calling protocols, approval gates, judge models, and guardrail frameworks.
• Evidence of technical contributions through open-source projects, publications, technical writing, or production systems.