Belmont Lavan is hiring an Agent Security Engineer to secure autonomous AI agents before and after they reach production. Agents now call tools, hold credentials and act on live systems, so they need the controls we already expect for people and services: identity, least-privilege access, monitoring and testing.
You will sit between AI engineering and security. You will design the guardrails, attack the systems yourself, and build the governance controls that let the business deploy agents with confidence.
Requirements
- 5+ years in security engineering (AppSec, cloud security, IAM or offensive security), with at least 1 year securing LLM or agent-based systems
- Hands-on knowledge of LLM attack techniques: prompt injection, jailbreaks, data exfiltration via tools, RAG poisoning
- Experience designing identity and access controls: OAuth 2.0 / OIDC, token exchange, workload identity, secrets management
- Working knowledge of agent frameworks and protocols such as LangGraph, OpenAI Agents SDK, CrewAI or Model Context Protocol (MCP)
- Strong coding skills in Python, plus one of Go, TypeScript or Rust
- Cloud security on AWS, Azure or GCP, including logging, detection and policy-as-code (e.g. OPA / Cedar)
- Ability to explain AI risk clearly to engineers, risk teams and executives
