Local-first AI
A question deserves a source.
GBC-AI
A sermon search for my church, with local transcription and answers linked back to the moment they came from.
Raymond Riter / Full-stack developer
I’m Raymond, a developer and dad in Medina, Ohio. I build local-first AI, tools for my family and church, and hands-on experiments that make complicated ideas click.

Interactive learning
Trace a sorting algorithm. Predict a swap. See why it works.
Start learningSelected work
Local-first AI
GBC-AI
A sermon search for my church, with local transcription and answers linked back to the moment they came from.
Open-source tools
Tesla OSS Suite
A connected workspace of Tesla tools, reproducible demos, and the decisions behind them.
Live on the course
GolfRay
A GPS caddie you install and walk the course with, built to work with no signal and shipped with no map tiles at all — and why.
Research in progress: Ten-Proof Frontier Studio, ten formalized results explored as software, with what remains unproved clearly laid out.
Choose your starting point
You do not need a polished specification or the right technical vocabulary. Start with the person, the problem, or the annoying part. I’ll help figure out whether a small app is worth trying.
01
Tell me who it helps and what you wish it did. A rough message is enough to start.
Tell me the idea02
Show me the repetitive, fragile, or expensive part. We can work backward from the real bottleneck.
Describe the bottleneck03
I build honest website concepts for nearby businesses without inventing facts or publishing before approval.
See the local website studioOhio teacher or principal? Explore practical AI help for your school
Just checking the work? Browse every project
My wife, my son, my church, and the occasional batch of homemade beef jerky. A lot of the things I build start right here.





An on-course GPS caddie built in fifty-nine verify-then-advance loops, and why the first public release makes zero map-tile requests: the imagery licenses would not allow the offline cache a golfer actually needs.
An interactive tutorial for four solving methods where the correctness guarantee is a test, not a library — and where that test caught a published algorithm that could not actually solve its own case.
An interactive lab that turns sorting algorithms into decisions you can predict, inspect, and test—not animations you simply watch.
I have a Model Y on order. I also have an RTX 3090 and a habit of building tools out of frustration. The result: 8 dependency-ordered Tesla repos, 6 of them live on PyPI, 611 passing tests, and a workspace that any owner can run without a vehicle in hand.
Project stories, things I learned, and the occasional smoker recipe. Read new posts in your favorite feed reader.
Subscribe via RSSPushed to rtk on .
View GitHub profile