I build production AI systems that work under constraints
that don't forgive sloppiness.
All production. All serving real users. All built solo.
Multi-persona AI executive assistant. 4 concurrent specialized agents with shared memory, autonomous operations, and 25+ scheduled heartbeat crons running overnight research, maintenance, and synthesis.
Autonomous AI fitness coach via Telegram. 5-tier deterministic intent routing saves ~70% of API costs. 12-tool agentic coaching loop with forced read-before-respond safety. Zero-LLM crisis handler for safety-critical paths.
Production voice AI handling live inbound staffing calls. Multi-step agentic workflow: speech-to-text → intent classification → database operations → response generation → text-to-speech. Automated screening, resume intake, job matching.
End-to-end staffing platform focused on recruiting. Admin portal, candidate pipeline, campaign management, AI-guided conversational onboarding with step-by-step Whippy integration setup.
I don't just build. I think about how to build.
Start with the most powerful AI model available. Progressively shift to cheaper and more deterministic solutions as the problem stabilizes. Let intelligence find the shape, then harden it.
Define. Delete. Ship. Measure. Trust + Release. Five steps. No sixth. Every system I build follows this cycle.
Structured context delivery for AI systems. The difference between an AI that knows what to do and one that asks.
Node, Lattice, Mesh — Beer's Viable System Model made nameable. One node type, two modes. The system doesn't change; your ability to see it does.
A business advisor built on Hormozi's published frameworks. Real vector search. Real retrieval scores. Real cost tracking.
Straight out of high school. Call center → grounds supervisor → warehouse manager. Learned leadership by doing the work nobody else wanted.
Electrician's Mate Nuclear. SSBN-733 USS Nevada. Qualified Engineering Watch Supervisor, QAI, N14S. Ranked #1 E-5 aboard. Used data analysis to drive training quality, enabling the boat to pass ORSE. Kept handwritten notebooks full of AI system architectures during off-watch hours.
Radiological Controls Technician. Led the team that certified the base for emergency nuclear operations. Planned and executed drills. Documented everything.
Separated September 2024. The submarine notebooks finally had a runtime. Started building immediately.
BS Data Analytics (WGU, 2026). 5 production AI systems in 14 months. BillyFit, IRIS, FrontLine, JobLink, ACQ Advisor. The throughline is systems — understanding them, building them, making them work under constraints that don't forgive sloppiness.