Kidus Y.

I build mobile apps end to end with Flutter, and machine-learning systems that hold up on real data. I've designed and built 25 mobile apps for Android and iOS, 24 of them in Flutter and 11 released, from calendars and reading apps to games, AI assistants and a camera. I take a product from idea to a signed release: the design, the code, the tests and the release pipeline. What I do on mobile: • Flutter and Dart apps for Android and iOS, with Riverpod or Provider • Custom UI: custom painters, motion and animation, home-screen widgets • Native Android in Kotlin through platform channels (Pigeon) • Offline-first storage (SQLite), local notifications and background work • AI features in apps: LLM tool calling, speech-to-text, on-device ML Kit • Backends: Supabase, Firebase and Cloudflare Workers • Localisation, including Amharic and the Ethiopian calendar • Google Play releases: signed App Bundles, CI with GitHub Actions, store listings and privacy policies What I do in machine learning: • Retrieval and RAG systems: embeddings, vector search, hybrid search and reranking • LLM apps with Azure OpenAI, OpenAI, Claude and Gemini, including agents and tool calling • Classic ML: LightGBM, XGBoost, CatBoost and scikit-learn, with honest validation • Computer vision with OpenCV and MediaPipe • Python and FastAPI services, Docker and Azure deployment A few things I've shipped: • Glass Calendar: glass home-screen widgets, the Ethiopian calendar and Amharic • Brain Dump: speak for 30 seconds and get tasks, reminders and notes, sorted by AI or offline • Nova: a voice assistant that makes calls and sets reminders and alarms, in Amharic and English • Loops: private search and Q&A over Gmail and Telegram on open models I care about apps that feel calm and considered, code that's tested, and clear communication. Tell me what you're building and I'll tell you how I'd approach it.