Projects
Software I built on my own time. I’m showing you the status of each one honestly, including the ones that are down right now, because a portfolio that only shows the wins isn’t telling you anything useful about how someone works.

Custavox
custavox.comCX intelligence for community banks and credit unions
Banks collect feedback constantly: reviews, surveys, complaints. Almost none of it is attributed to anything. Custavox connects each piece of feedback to organizational structure, meaning a branch, a department, an individual, a system, or a touchpoint, so a pattern becomes something a specific person can act on.
It is built for community banks and credit unions under $1B in assets, institutions that rarely justify a full-time Chief Experience Officer. Custavox is positioned as that function, fractional and software-driven. I founded it and I build it, alone. There are no customer results to report yet and I'm not going to pretend otherwise. This page changes when that changes.
- Claude Code, Vercel, GitHub, Supabase, Anthropic API
- In progress. Phase 1 complete, in discovery with prospective community banks.

Vizzls
Offline while I rebuildAn AI-built personal site for job seekers
A resume is a list. A personal site is a case. A job seeker uploads a resume, and Vizzls parses it into a personal one-page site with narrative copy, across multiple templates. A LinkedIn-style recommendations feature lets former colleagues and managers vouch for the person directly on their site, where a hiring manager will actually see it.
The site you are reading is the same idea, done by hand. Vizzls exists to give other people the same thing without building it themselves.
- Claude Code, Vercel, GitHub, Supabase, Anthropic API
- Offline while I rebuild the template engine. The recommendations flow is the part I'd keep.

PiggieFly
Offline, on the shelfLicensing requirements as a step-by-step roadmap
Every state and industry has its own licensing requirements, and a first-time founder has to assemble them from many sources. Running IASourceLink meant answering that question over and over. PiggieFly is the answer, productized: fragmented state and industry requirements turned into a personalized, plain-language roadmap. I built it solo with no-code and low-code tools, which kept the work focused on the content and the logic rather than on infrastructure.
- Replit, Anthropic API
- Offline. Built on no-code tooling I no longer maintain. The licensing logic underneath it still holds up and would be worth rebuilding.

Ghouler
Offline, rebuildingHaunted locations as structured case files
Most haunted-location content is entertainment. Ghouler treats each reportedly haunted U.S. location as a case file: historical background, reported phenomena, and sourced documentation, aggregated from folklore, local records, and investigative material. I built and launched it solo, with an information architecture designed to hold thousands of entries without the structure degrading. The same discipline as a licensing database, applied to a very different subject.
- Replit, Anthropic API
- Offline while I move it from a prerendered pipeline to Next.js with static generation. I traced a total Google indexing collapse across 9,000 pages to snapshots being taken before content loaded, which produced empty HTML shells that got classified as soft 404s. Fixing that is the rebuild.