I build AI-powered operations software for US school districts, from Lima.
Full-stack product engineer. I take systems from spreadsheet prototype to production — and I’m the one who gets called when the numbers stop reconciling.
Guardrails that stop AI-written code from shipping production bugs. Four guards — timezone safety, money math, tenant isolation, migration safety — each written from a failure that actually reached production, not from a style guide. Installs as Claude Code skills with a /guard pre-ship sweep.
Claude Code · Bash · PowerShell · GitHub Actions · MIT
Texas rates every school A–F, and districts buy tutoring to move that grade. Sales was hand-building each proposal over two days. I took the model from spreadsheet prototype to production platform and rewrote the optimization until it reached the same target grade on 49% less budget — which is what made the proposals win. Prep time went from two days to two hours.
React · Node.js · PostgreSQL
Leading five engineers on the system that delivers those contracts: tutor matching across availability, language and grade level; session lifecycle with cascade rescheduling; attendance-driven billing; and the monitoring that catches a missed session before the school does. 51 schools, 900+ tutors, 2,700+ students.
Next.js · NestJS · PostgreSQL · AWS
I started in product and moved into engineering, which means I argue about the requirement before I argue about the implementation. Most of the expensive bugs I have shipped were not typos — they were a faithful implementation of something nobody had thought through.
I write production code with AI agents every day. The hard part is not generation, it is verification — specs an agent can be held to, adversarial review passes, and CI guards that turn a fixed bug into a rule the codebase enforces on its own. angelstack is that discipline, extracted and made installable.
davidchcc1118@gmail.com — I read everything and reply to anything specific.