A LoRa mesh across Phoenix
Deploying MeshCore nodes through the West Valley. Traced a board that received fine but never transmitted down to an antenna-switch bring-up bug in the I/O expander — fixed upstream in a community fork.
Read
Off the clock
Making
Things I build because I want to know how they work. Radios, meshes, models running on hardware that has no business running them. Nothing here was assigned.
When to point an antenna at the sky
A pass planner for receiving satellites with a $30 software-defined radio. It predicts when each one is overhead and hands you everything needed to capture it: frequency, antenna length, radio settings, and the decode commands.
Read Code ↗
One model across two laptops
Pooling the GPUs of a Windows laptop and a Kali laptop over Ethernet to load a language model that fits on neither. It works, it's slower than you'd hope, and measuring exactly why was the point.
Read Code ↗
A flight tracker that fits in a pocket
Firmware for a card-sized ESP32 computer that shows live air traffic as a radar scope, a sortable list, and a detail card per aircraft — with a watchlist, logging to an SD card, and proximity alerts.
Read Code ↗
Aircraft, from the desk
An RTL-SDR feeding readsb and tar1090 on Linux, plus a Flightradar24 feed. Built a V-dipole tuned for 1090 MHz and calibrated it against known traffic.
Read
Language models on six gigabytes
Running quantized models on consumer VRAM — working out what Q4_K_M actually costs in quality, and reading enough of nanoGPT to know why attention works.
Read
Contact
Let's build something.
I'm looking for data science, machine learning and AI engineering roles in Phoenix or remote — and I'm always happy to talk about models, retrieval, or why yours is overfitting.