I Shipped a Chrome Extension in a Weekend. I’d Never Built One.
An ML engineer, an on-device AI tool, and how little the not-knowing cost.
Lens: Explaining Any Word on the Page, On-Device
An ML engineer with no web-dev background ships a Chrome extension in a weekend with Claude. What it does, what I learned about the browser platform along the way, and what the on-device model's latency forced the design to become.
I Built a Fully Local Voice Assistant on a Raspberry Pi Cluster. The Hardest Part Wasn’t the LLM.
A privacy-first Alexa-style assistant running on Raspberry Pis, where the real lesson turned out to be latency, tool design, and wake-word…
Building a Fully-Local Voice Assistant on a Raspberry Pi Cluster
How I built an Alexa-equivalent that runs on three Raspberry Pis (a fourth handles storage), and what a wake word that kept firing on "ok" taught me about training data
The Meta-Lesson: ML Infrastructure Is Mostly Software Engineering
One truth I continually relearn is that building machine learning infrastructure is 80% software engineering and only 20% machine learning…