Contributions
Every problem I hit, I follow with a fix.
Most of my current work is open-source contribution to Lemonade and the tools around it — making local AI run reliably on the hardware people already own. The pull requests are the most honest record of it.
Where the work lands
Four places I keep showing up.
Broad hardware support
Getting AMD graphics cards to work as reliably as Nvidia — detecting the right compute path automatically so local AI runs on the hardware you already own.
ROCm & Vulkan enablementReliable model management
Fixing the bugs that show up when you load or switch models, so downloads and status reporting are accurate and you never have to guess whether a model is ready.
Model lifecycle & health checksEasy-to-use developer tools
The command-line tools and launchers that start the work — made to behave the same on Windows and Linux, removing the setup friction.
CLI ergonomics & cross-platform fixesSettings that stay put
Remembering preferences like themes and filters, kept locally on your device without getting in the way of the AI engine itself.
Local state & UI persistenceI write about a lot of this as I go — the Local AI Stack series walks from "what is my GPU doing" all the way to opening your first pull request.