Client work

Real systems, shipped for real businesses.

A short tour of what I've built recently. Each one started with a business with a big problem and ended an effective solution.

Building a ranked lawyer database from public records.

The firm's sales team needed custom email and phone lists of personal injury lawyers, ranked by the number of cases each attorney actually handled. Off-the-shelf legal directories don't cut it: they don't tell you who is busy and who is not.

Scraped the Arizona state bar to build a roster of every licensed attorney.
Pulled civil case records from the state's largest county court database.
Classified every attorney by practice area and joined the data, so the team could rank by litigation volume.
200,000 attorneys profiled. 150,000 court cases collected. Arizona covered end to end.

Removing six hours a day of manual listing work.

The marketing team was spending six hours a day editing thousands of online listings by hand. House plans were scattered across folders and email threads, with no single place to find the current version of anything.

Wrote a Playwright automation that posts new listings to Zillow end to end.
Built a database tool that organizes home plans in one place, with version history.
Over 2,000 listings posted through the automation. Listing work went from a full afternoon to a few minutes.

Taking over a RAG system and making engineering decisions defensible.

The parent company had hired an outside firm to build the first version of an AI RAG system, a spinoff product alongside their core catering business. The build shipped, but they needed someone to take over the codebase, keep it running, and ship iterations as real customers started using it.

Took ownership of the codebase from the original development firm.
Maintain the system and ship improvements as feedback comes in.
Built measurement tools that show which prompts performed and where the model was costing money without adding accuracy.
Replaced guesswork with a quantitative basis for engineering decisions.

Putting a local golf shop online with smart pricing built in.

A Maryland golf store wanted to start selling online. They also needed a way to price used clubs against the live market instead of guessing.

Designed and built the online store from mockups through launch.
Built custom pricing tools that pull live market data to price clubs.
A single-owner shop now selling online with margins they can defend.