22 years keeping the U.S. Navy connected. Now I lead a team putting modern AI on ships and in air-gapped networks — where it has to just work, with no reach-back and nobody to call.
I joined the Navy in 1999 and spent the next twenty-two years making sure the fleet could talk — on a patrol craft, a destroyer, an aircraft carrier, at a comms station in Hawaii, at Fort Meade, and downrange in Iraq and Afghanistan. I retired as an IT1 in 2021, but I never really left the mission.
Today I'm the contractor lead and AI program coordinator for the CB-ISEA Fleet AI Modernization Program at NIWC Atlantic. My job is to take the AI everyone's excited about and make it real on a ship — on hardware we can actually install, on networks that will never touch the internet, for technicians who need an answer right now.
I'm entirely self-taught. I'm the guy with a GPU server humming in the house and three side projects going at once, because I genuinely can't stop tinkering. If you're building something hard, I'd love to hear about it.
Every stop taught me something I still use: how ships actually run, how to lead people under pressure, and how to keep systems up when nobody can fly out to fix them.
Leading a contractor team driving Navy AI modernization — air-gapped deployments, GPU-optimized inference infrastructure, and getting AI tools into the hands of fleet technicians.
Kept Linux servers hardened and running, owned vulnerability scanning and remediation, and managed a COMSEC account — the kind of job where "close enough" isn't an option.
Directed a team of 7–10 subject-matter experts supporting the fleet's CANES network across five software baselines — installs, system verification, troubleshooting, and the Kubernetes underneath it all.
Three ship tours (USS SIROCCO, USS JAMES E. WILLIAMS, USS GEORGE WASHINGTON). Fleet messaging at NCTAMS PAC with an Iraq deployment. NIOC Maryland at Fort Meade with an Afghanistan deployment. Learned that the network is the mission.
Most AI assumes a fast connection to someone else's cloud. A ship in the middle of the ocean doesn't get that. That gap is where I live.
Self-hosted language models and retrieval systems that run fully offline — on hardware sized for a server rack, a lab, or a ruggedized case.
Spec'ing, building, and tuning GPU servers for fast local inference. If it has a fan and a CUDA core, I've probably had it apart.
My open-source lab for pitting language models against humans at chess — and finding out which AI can actually play.
View on GitHub →Before the AI boom, while I was supporting the fleet's network, I took an experimental project off GitHub and turned it into a retrieval-augmented AI on a server in my house — loaded with about 1,100 troubleshooting documents. I put it behind a VPN so I could ask it questions from my phone while I was standing in a ship's server room. That's the moment I knew where my career was going.
I've been the sailor at 0300 with a down system and no help coming. Everything I build starts from that person.
I'm not the person writing every line of code — I'm the one who knows what the system must do, directs the team (people and AI tools alike), and doesn't stop until it ships.
I'd rather give you an honest "here's what could break" than an exciting promise. Overclaiming gets people hurt in this business.
Leading sailors and SMEs for two decades taught me that teams win when people know you've got their back.
Tap a question. These are the things people usually ask me over coffee.
Serious collection. The GitHub handle "bobathefetts" should tell you who my favorite is.
Reverse-engineered an RFID protocol and printed a custom enclosure — just to build a theme-park gadget for fun.
Long-running games with a coworker, played on an offline board I built. And yes, NeuralChess started here.
Anything that flies and has a radio in it. Old habits.
High-end rigs, GPU servers, and the occasional crypto-mining experiment on a Raspberry Pi cluster.
ARC Raiders and Call of Duty when the kids are asleep.
Theme parks and cruise ships with the family whenever we can swing it.
Which means I know a thing or two about patience and resilience.
Whether it's a role, a project, or just swapping notes on running AI offline — reach out. I answer.