Building a FastAPI endpoint for churn prediction, and everything that broke between "it runs" and "it's live." The project started as a small FastAPI app to predict customer churn, returning a probability, prediction, and risk level. After testing it locally in Swagger UI, the author realized the model was useless unless it could survive a sleeping laptop and reach strangers online. The fix meant containerizing the app and moving it to a real server.

Docker solved environment mismatches by bundling Python 3.14.6, specific package versions, and file paths into a self-sufficient image. The build took 166 seconds because some packages compiled from source, and three separate failures appeared only after deployment began.