Founding Agentic AI Engineer

About the job

GuardiumOne is next generation autonomous monitoring infrastructure using agentic AI. As Lead Agentic AI Engineer on the Policy Intelligence team, you'll design and own the agent graph systems that continuously monitor our platform for policy violations, compliance gaps, and configuration drift — and take automated action to close them. This is a high-ownership, high-autonomy role at the intersection of LLM systems, graph reasoning, and security engineering.

Responsibilities

• Architect and build multi-agent graph pipelines (LangGraph, custom orchestration) that monitor IAM policies, Kubernetes RBAC configurations, and data access controls across the platform in real time.

• Design semantic policy gap detection — building LLM-powered reasoning layers over OPA/Rego rule sets, NIST controls, and internal policy baselines to surface ambiguous or uncovered cases that rules alone miss.

• Own the end-to-end observability stack for agent execution — tracing node state, tool calls, and decision branches across graph runs to enable fast debugging and drift detection in production.

• Close the loop on detected gaps — generating structured remediation outputs (PRs, Jira tickets, policy patches) that feed downstream enforcement pipelines with minimal human-in-the-loop intervention.

• Define evaluation frameworks and evals for LLM-based policy reasoning — measuring precision, recall, and false-positive rates on gap detection across compliance domains (SOC2, ISO 27001, internal SLOs).

• Mentor 2–3 ML engineers on the team; contribute to architecture decisions, code reviews, and the ML platform roadmap as a senior IC voice.

Required:

• 7+ years in ML engineering with 2+ years shipping multi-agent or agentic LLM systems to production

• Deep hands-on experience with LangGraph, LangChain, or equivalent graph-based agent orchestration

• Working knowledge of graph databases (Neo4j, Memgraph, Amazon Neptune) and knowledge graph construction

• Familiarity with policy-as-code tooling — OPA, Rego, or equivalent rule engine frameworks

• Strong Python; experience with FastAPI or similar for serving agent endpoints

• Proven ability to design and run agent evals — not just vibe-checking outputs

• MS or Phd in Computer or Data Science

• Excellent verbal and written communication skills