Threat-model AI agent features for prompt injection, jailbreaks, data exfiltration, and over-privileged tool use.
Design guardrails including least-privilege scopes, human approvals, output filtering, and tenant/data isolation.
Secure the AI supply chain, including models, providers, MCP servers, connectors, prompts, and configuration.
Run AI red-team exercises and build automated adversarial tests into CI.
Define Kubernetes and cloud security baselines covering RBAC, network policies, workload identity, secrets, and tenant isolation.
Implement CI/CD security controls such as IaC scanning, SAST, dependency scanning, and policy-as-code.
Build detection and alerting for anomalous prompts, tool calls, and agent behavior.
Partner on technical controls and evidence for SOC 2 Type II, ISO 27001, HIPAA, ISO 42001, and the NIST AI RMF.
Requirements
5+ years of experience in cloud security, application security, or DevSecOps.
Deep hands-on Kubernetes and cloud security experience, with GCP or Azure preferred.
Security automation experience with Terraform, Helm, and Python or Bash.
Working understanding of LLM application risks, including the OWASP Top 10 for LLM Applications and agentic system failure modes.
Ability to influence engineers without formal authority.
Preferred experience securing LLM, RAG, or agentic systems in production.
Preferred experience with AI red teaming, MITRE ATLAS, CNAPP/CSPM tooling, SIEM or detection engineering, AI security research, open-source tools, or CTFs.
Benefits
Work on production agentic AI systems and make security a core product feature.
Partner directly with Security, Compliance, Engineering, and Product leadership with substantial ownership.
Individual-contributor culture focused on execution and impact.
Compensation may include variable compensation, equity, and benefits depending on role eligibility.
Ema
Ema is a suite of AI employees for HR, IT and Finance. She doesn't just automate workflows — she redesigns the work around your outcome and executes it across every enterprise system. Proven at scale across the world's largest enterprises.