Train and Serve a Weather Prediction Model

In Project 1 you built the pipeline. Now you do what ML engineers do with pipeline data: turn it into a model that makes a prediction, prove the prediction is actually worth something, and put it behind an API that anything can call.

SQLite (Project 1)  →  features.py  →  time-aware train/test  →  train.py  →  model.joblib  →  serve.py (FastAPI)  →  GET /predict

What you'll build

A model that predicts tomorrow's mean temperature for your three cities, trained on the data your own pipeline collected. You'll engineer time-series features without leaking the future, beat a "dumb" persistence baseline (the bar every real model must clear), save the winner with joblib, and serve it from a FastAPI /predict endpoint you can hit with curl.

How this works

Work through the steps in order. Lessons teach the concept. Quizzes check you got it. Milestones are where you build — you'll paste your code and output, and your AI mentor reviews it like a senior engineer would in a pull request: what's good, what's wrong, and hints (never the answer).

Prerequisites

Project 1 (Build a Weather Data Pipeline) — you need its SQLite database and its code. If you've only run the pipeline for a few days, no problem: step 1 shows you how to backfill ~3 months of history with one parameter change (past_days=92). You should be comfortable with pandas basics from Project 1; scikit-learn and FastAPI are introduced when you need them.