Builder / operator. I turn ambiguous problems into working systems, across government R&D, small business, and applied AI.
I'm a materials engineer by training (Ph.D., Purdue) and a builder by habit. For the last seven years I've led an advanced materials laboratory at Naval Surface Warfare Center Crane. I also own and operate two small businesses (one with partners) in Bloomington, Indiana, and I spend a lot of my spare time building AI agent systems that actually get used.
The common thread is the same loop every time: define the problem, build the thing, test it honestly, find out why it broke, and decide when the evidence is good enough to ship.
Stood up a ~$2M advanced materials laboratory end to end, and lead research programs from idea to proof of concept.
A coffee roaster and a café, refillery, and community classroom, including the systems and tools that run them.
A personal agent-based operations environment with a "second brain," guarded permissions, and AI coding agents that need approval to ship.
$ curl https://www.matthewmichie.com/.well-known/agent-card.json { "name": "Matthew Michie", "version": "1.0.0", "runtime": "biological", "requires_gpu": false, "requires_coffee": true, "human_in_the_loop": true, "status": "available_for_deployment" ... full card }
A human agent card in the A2A 1.0 format, published at the standard well-known path. It's an Easter egg, not an instruction: decisions about humans belong to humans.