The work you can't see on GitHub
The open source on this site is the part I can show you. It is real, and it is what I point to when someone asks how I work. But it is not most of what I have built. Most of that runs inside companies, under NDA, for people who will never know my name.
I cannot show you the code. I can tell you what the problems were, which is the part worth reading anyway.
Integration at scale
A hospitality client ran hotels, casinos, and resorts, and needed their point-of-sale systems to talk to around 70 loyalty and rewards platforms. I built the API gateway that sits in the middle, plus 25+ partner adapters that run on an embedded JavaScript engine so a new partner can be added without redeploying the core.
The reason it had to be right: a wrong points calculation is not a failing test, it is a wrong charge on a real guest's bill.
Down to the hardware
A time-and-attendance platform used by more than 10,000 schools and organizations, processing millions of sign-ins a week. I worked on the desktop software and the school-system integrations, and I tested the physical kiosks and RFID terminals from the screen a student taps down to the row in the database.
When software decides "was my kid at school today," it has to be right at every layer, including the ones with screws in them.
AI features in production
A learning platform I owned from design through deployment. Figma to the finished front end, SCORM e-learning, Stripe payments, and LLM features that curate content and narrate articles using local models. Every release runs through a CI pipeline with end-to-end, load, and security checks before it goes out, because an AI feature behaving strangely in front of paying users is a bad afternoon.
Software that catches a lie
A monitoring system for farmland that tracked whether each plot was actually prepped, planted, and harvested. The real problem turned out to have nothing to do with farming. Inspectors were signing off on land they had never walked, using a photo taken somewhere convenient.
So the app captures every photo with its GPS coordinates and navigates the inspector to the exact plot, which can be a long way out. The photo now proves where the person was standing, not just that a photo exists.
Others I built, or was handed broken
- A billing summary that used to take hours now runs in about five minutes. I rewrote the SQL first, then reworked the code around it.
- A records and medical laboratory system: lab results, patient records, billing, and the daily cashflow the owners run the business on.
- A sports analytics app that was too laggy to use. I profiled it, fixed the code, and it runs smoothly now.
- A drilling company's legacy API, moved onto PostgreSQL and .NET Core while it stayed in service.
- A school was stuck with a closed-source system that did not do what they needed and came with no documentation. I built something that integrates with it anyway.
- A census app for community surveys: who lives where, household size, economic standing, voter records.
- A hotel management system with the daily ledger and cashflow the owners close the day on, wired through to the point of sale.
- A consumer electronics site that could not properly store or expose its own product catalogue. I rebuilt the backend so it could.
If any of this sounds like a problem you have, email me.