What does it actually take to ship an AI feature in a Flutter app? Not a demo, but something that runs in production. This talk follows modemoiselle, a Flutter and Firebase wardrobe app, from the first Gemini call to a live product, and focuses on the lessons that cost real time and money. Where AI belongs in the architecture and why it lives in exactly two files. How structured output turns a photograph into typed data you can trust, and what happens when the model is slow, wrong, or unavailable. Then the unglamorous half: retries and typed timeouts, latency you cannot hide behind a spinner, a CORS failure that reports itself as status code 0, and a 10 EUR budget alarm that taught me an alert is not a cap. You will leave with an architecture blueprint for Flutter plus Gemini, honest cost and latency numbers from a running app, and patterns you can reuse the next day.

Audience levels
Beginner-Intermediate- Advanced (adjustable)

Session topics
AI - Antigravity - Gemini-Dart - Flutter- Firebase