I build the production plumbing that lets LLMs operate real systems. Retrieval, agent tool-calling, evals, and the identity layer that keeps an agent from acting as a god-mode service account.
Currently a Senior Automation Engineer at an MSP, where I'm the sole engineer on two production AI platforms: an LLM ticket-triage service, and a hosted MCP server exposing ~105 tools across nine vendor platforms behind an OAuth 2.1 authorization server with privilege-checked on-behalf-of identity.
Before that: eleven years of systems and automation work, and two years in the Navy as a nuclear propulsion machinist's mate, which is where I learned that the interesting question is always what happens when it fails.
Working on: agent trust boundaries. Tiered autonomy, approval gating, and how you give an agent increasing authority without giving it all of it at once.
Tools: Python · Rust · PowerShell · PostgreSQL · Kubernetes · MCP · LangGraph
Most of my work is proprietary and lives behind an employer, so this profile is thinner than my resume. Happy to talk through any of it.