Module 2 · Project 1: Serve AI Models as an API (FastAPI)

OBJECTIVE: turn two AI models into a proper web API — a classic ML model (credit-risk prediction) and an LLM app (a document Q&A assistant with streaming answers) — with input checking, documentation, tests, security basics and a simple web front end. WHAT WE COVER (in project order): • The starter models (reuse yours or use the provided small ones) • FastAPI basics: routes, request/response models with Pydantic, automatic Swagger docs • Serving an ML model: load once, validate input, predict, version • Serving an LLM app: async calls, streaming responses, timeouts and errors • Error handling and structured logging • API keys, CORS and rate limiting • Automated tests with pytest • Front end: Streamlit (and Gradio) calling your API AFTER THIS MODULE YOU CAN: • Wrap any model or LLM pipeline in a clean, documented, tested API • Stream LLM answers to a front end • Protect an API with keys and rate limits TOOLS: FastAPI, Pydantic, Uvicorn, pytest, Postman, Streamlit/Gradio.