Built end-to-end a cross-platform desktop app (Tauri 2 + React + TypeScript) that captures screen text via OCR (Tesseract.js), supports meeting-style transcripts (Web Speech + manual input), and answers questions with a local LLM (Ollama) grounded in that context — with explicit consent, visible “capture on” state, and SQLite storage. Implemented the data layer, context engine, session history, export, settings, error handling, Vitest coverage, ESLint, and CI (including Rust checks); ran release-hardening, documentation, and macOS bundling workflow.