AI-Rey Field Notes
Neural networks, field systems, and the engineering behind useful AI. Technical writing by Roberto Vega, with a public trail of work from 2016 and 2018.
Start with the archives. Follow the decisions. Bring the useful parts into your business.
Embedded AI & remote sensing
An ESP32-CAM, a roughly 4 MB quantized YOLO model, and a demonstrated path from live classification through LTE to a native observer console.
Read the article →Neural networks
From the 2016 spelling experiment to exact labels in 2026: three labels at once for “What is my name?” and 9 exact matches on 12 sentence-type checks.
Read the article →AI in the physical world
On unseen storms, my ANN’s peak-elevation difference was 1.31 feet one hour ahead and 0.72 feet at timing, compared with 1.08 to 1.95 feet for the reference models at timing.
Read the article →AI for business
A practical technical guide to choosing a bounded text-classification task and proving that its output is useful.
Read the article →Put a useful workflow to work
AI-Rey implements agents, connects supported tools, and trains the people using them. Start with one routine task and a useful result.
