01
AI builds, end-to-end
Strategy, model selection, data design, and the actual code that ships. I work the whole stack so the proof-of-concept doesn't die in a Jira backlog.
ABOUT
I run a small company that builds AI tools for credit unions. The job is half engineering — picking models, designing data, writing the actual product — and half listening for the part of someone's week they hate most, then quietly making that part disappear.
Most of the AI work I'm proud of is the unsexy variety. I'm less interested in agents that simulate humans and more interested in tools that remove the meetings nobody wanted to be in.
WHAT I WORK ON
02
Internal tools and automation
Teams without a full engineering org still have repetitive work that's quietly eating them. I find that work and write the tool that makes it not.
03
Coaching for leaders and teams
One-on-one and small-group workshops. Help cutting through the AI hype, picking what's worth building yourself, and getting a team to actually use the thing you bought.
THINGS I'VE BUILT
01
A CRM for credit unions, built on call-report data
Sits on top of NCUA 5300 quarterly call report data. Lets credit unions find the right partners, and lets fintechs find the right credit unions. Five user roles, each with their own slice of the same dataset — watchlists, intent-signal scoring, M&A tracking, and AI-assisted strategic planning. The whole thing runs on a unified AI wrapper with usage metering underneath, which keeps the bill honest.
- A spinoff data API. The same NCUA dataset, repackaged as a standalone subscription product for external buyers. API keys, tiered rate limits, Stripe billing, marketing site.
- A contact-email predictor. Predicts likely emails from name + domain patterns, runs them through verification, and surfaces a diff preview so importers can confirm matches before anything writes to the CRM.
Next.jsPrismaPostgresTypeScriptHonoStripeAzure
02
An internal operating system
Where most of our company actually runs. A FastAPI service that handles proposals (with a PDF library and a prospect-facing UX), client onboarding flows, deal tracking, and AI coaching layered over call activity. It's the system of record for the pipeline — everything else pushes events into it.
PythonFastAPIPostgres
03
An AI meeting copilot
A bot service joins calls, transcribes them, and produces structured summaries that route automatically into the right downstream system. A Next.js + Prisma app handles the routing logic, AI summary refinement, and human-in-the-loop confirmation, with undo support and reprocess progress baked in. Closest thing I've built to a real-time agent.
Next.jsPrismaTypeScriptAI agents
04
A scheduling tool for B2B sales
Calendly-style, but built directly on Microsoft 365 so it slots into how most of our customers already work. Multi-tenant Azure AD sign-in, host availability driven by O365 calendars, and a public booking page that drops a Teams meeting on the host's calendar. pnpm monorepo on Azure Container Apps. Recent versions added lunch breaks, half-day blackouts, an approval workflow, and manual bookings.
Next.jsTypeScriptMicrosoft GraphAzure
OFF THE CLOCK
Piano & guitar
Mostly by ear. Progress is uneven but compounding.
Game development
Side projects in the space where storytelling and systems design overlap.
Gardening
Slow returns, deep satisfaction.
Rivers
Kayaks, maps, and the kind of curiosity that costs gas money.
Snowboarding
The cheapest way to feel like you're flying.
Old books
Philosophy, esoteric texts, and the occasional manual.
Starting things
The throughline, for better and for worse.