On-device AI & ML systems — I build tools that run locally, not in someone else's cloud. 🦉 Night strategist.
linux-lightweight-stt — push-to-talk dictation for Linux that runs Whisper on the Intel NPU instead of the CPU or a cloud API. Hold a key, talk, let go, and the text lands at your cursor. Wayland-native injection, ~250 MB, near-zero idle cost.
Also in the workshop: RAG pipelines, agent tooling, and Rust for the parts that have to be fast.
