I'm not an engineer by trade. I have a non-technical day job and taught myself to build by shipping. These days I make a steady stream of small AI products and developer tools, mostly with AI-assisted workflows, and I write up what I learn along the way.
The best known is EnviousWispr, an open-source (GPLv3) voice-to-text app for macOS that records, transcribes, and polishes speech fully on-device. Free, no account, private by architecture. I also run EnviousStaging, a virtual staging SaaS for real estate that generates furniture and decor for listing photos with a human in the loop. Around those are other Envious Labs products (image and video generation, social and automation tooling) and a pile of open-source CLIs and MCP servers on the side: a multi-model "LLM council," a CLI generator, macOS accessibility automation, and more.
What I like sharing here: the on-device ML experiments that worked, the ones that didn't, and the measurements behind both.