Building an AI chatbot that looks great in a demo takes an afternoon. Making one that customers can trust takes engineering. Most AI features fail the same way: they sound confident, invent answers, miss the right document, or quietly get worse after a harmless-looking change, and nobody notices until a customer does. In this 45-minute live session, we build a help-centre assistant for a fictional e-commerce company (ShopEase) on Claude, break it in front of you, measure how badly it fails, fix it, and then set up an automatic quality gate that blocks bad releases before they ship.

By the end of the session, you will understand why AI demos fail in production, how to measure an AI app’s quality with a golden set and an AI judge, and how to stop a bad release automatically with an eval gate - the same build-break-measure-gate loop AI Engineers use on the job. You’ll leave with a working repo and a clear picture of what it takes to become an AI Engineer.