Six case studies, the websites I've built for local businesses, and a few things for my family and neighbors. Case studies 02 and 03 are client work through Robots & Bacon, my part-time practice.
§ 016 case studies
Case studies
Case 01iotaMotion · 2016–2024
Taking a surgical robot from founding to FDA De Novo
Led product development from company start through FDA De Novo authorization in 2021, the first-in-human trial and early commercialization.
Why it matters in healthcare
A first-of-its-kind Class II device, taken through De Novo under design controls.
talking points per physician, each with a source link and a confidence level
Case 02A medical device client · 2026
AI research where every claim has a source
Pulls specialist physicians from public directories, researches each one and drafts ranked talking points, each with a source link and a confidence level. A person reviews before anything is sent.
Why it matters in healthcare
Traceability: every claim tied to its source, the same discipline as a design history file. The same pattern works for KOL mapping and market research.
Reps edit event proposals in plain English. Guards in code run before anything reaches the screen, so a price can only be a number the rep typed or "TBD".
Why it matters in healthcare
It's the core problem of AI in regulated work: let the model help, but make certain mistakes impossible, not just unlikely.
tables in one Postgres database, fed by 30 scheduled jobs
Case 04Party On Delivery · 2025–2026
One data platform with AI analysts on top
Grew the storefront into one Next.js app on one Postgres database with 30 scheduled jobs. Detectors and Claude Code analyst agents recommend; a person approves anything that changes.
Why it matters in healthcare
Read-only AI tools, separate access levels and a log of every request is the access-control and audit-trail pattern any patient-data system needs; the 14-day re-check is post-market surveillance for recommendations.
I built Wayne, the AI receptionist for my event-delivery business. Web chat, the phone line and texts all answer from one playbook of sourced facts, and the compliance guards fail closed.
Why it matters in healthcare
Patient intake and triage lines have the same shape: answer only from approved content, send anything outside it to a person, and let staff take over a conversation cleanly. The playbook works like document control: one approved source where every fact cites where it came from, generated copies, and a check that blocks drift.
A daily brief, shared memory across 12 projects and 25+ custom Claude Code skills and review agents. The agents can read, never send; every long run ends with an honest "Verified / NOT verified" list.
Why it matters in healthcare
Human approval before any change, a log of what was done and why, and independent review before release: design controls, applied to AI.
Boat and RV storage site with an interactive yard map and space finder. The same codebase and booking system also runs its sister site, 188 Self Storage.
Things I've built for my family and neighbors. Most started as "can I make this before bedtime?"
Fig. 1 The Fun House — 10 games I built for my kids with Claude Code.
The Fun House
A camp-journal website of ten games for my kids — Dolphin Dash, Equation Quest, Emoji Guess and seven more. Each is a single file that works offline. The hardest Emoji Guess puzzles were written by five AI writers, tested by three blind AI solvers and picked by an AI judge.
18
privacy rules enforced by the database
Neighborhood tool share
An invite-only phone app where the dads on our street lend tools: two-sided loan confirmation and 18 database-enforced privacy rules, built in two days.