Software built around real work.

Owner interviews, barcode lookup across five databases, checkout delivery checks, and a Supabase backend.

A pickup and delivery app for a convenience store near Miami's campus. Over several months I went to the store and asked the owner how he actually runs things, then built around his answers. Staff scan a barcode and the app checks his catalog, or pulls details for a new item from five product databases. Shoppers find their address with Google, and checkout confirms they're inside the delivery area. I imported his 2,219-row product sheet.

The first version, in 2025, is where I learned how the pieces fit together: Google sign-in, a QR pickup scanner, a live order map. This year I rebuilt the backend on Supabase. It hasn't launched yet and has never taken a real payment. The screenshot uses demo data.

React Native / Supabase / Google Places
2024-2026 · Built with AI coding agents

I own requirements, data checks, scheduled jobs, testing and releases. Claude Code and Codex write the code.

I wanted one place to research a game before betting on it. Parl.ai pulls every sportsbook's price, writes an AI preview of each game, and posts one pick a day. Every pick since January stays on the public record, losses included, and it never claims an edge. It's live on the web and the App Store, and I run everything behind it: the database, the scheduled jobs that pull odds and grade picks, and the iOS releases.

View on the App Store ↗

React / TypeScript / Supabase / Capacitor
2025-present · Built with AI coding agents

Started by learning how a friend in valve sales picks accounts. The hospital version uses public CMS provider data.

A friend in industrial valve sales told me how he decides which companies to call, and I turned that into a map of his territory. Then I reused it as a template for eight more industry maps built from public data. The one pictured ranks 733 US hospitals from CMS provider files.

JavaScript / SVG / Public data
2026 · Built with AI coding agents

Zoom into smaller fields, try hands-on discoveries, and earn color by answering a new example on your own.

This started as my senior capstone at Miami University. The idea is that learning should feel like exploring a world, not scrolling a course list. You zoom from eight big territories into smaller fields and try hands-on discoveries, and the map only gains color when you get a fresh example right on your own.

React / TypeScript / PixiJS / D3
2025-2026 · Built with AI coding agents

Find bill sponsors, committees and actions; search federal notices and explore Census data. Saves locally without an account.

I built this for a friend who works on Capitol Hill. Type a bill like H.R. 142 and it pulls the official record from GovInfo: sponsor, committees, actions, and summary. It also searches the Federal Register and maps Census data for every state. Everything saves in your own browser, with no account.

Next.js / TypeScript / D3
2026 · Built with AI coding agents

A side project for fun, with automated checks and desktop and phone reviews before releases.

A side project for fun: a made-up sports league that plays itself. Six sports, 48 invented clubs, and 416 pixel athletes, with games simulated live in the browser. Spin the globe, pick a city, and a match is already on. Every change had to pass the same checklist before it shipped: automated tests, desktop and phone screenshots, and zero console errors.

JavaScript / Three.js / Vercel
2026 · Built with AI coding agents