Building a Model Prediction Server
summary
Starts from a trained scikit-learn model (an Iris classifier the uploader notes as 100% accuracy), loads the pickle via importlib.resources, and embeds it in a small FastAPI app that exposes a status route and a POST /predict endpoint. Input/output schemas use pydantic models, pytest fixtures test the happy paths, and the app serves real-time predictions (e.g. setosa, versicolor, virginica) with automatic OpenAPI docs; the demo runs locally and the model actually returns plausible labels.
AI-written summary · about
chapters (6)
- 0:00Welcome
- 3:56Project workspace setup
- 28:59Simple FastAPI app
- 54:20Pydantic data models
- 1:11:07Scikit-learn model integration
- 1:33:36Prediction server finalized
related talks
01h 18mBuild Data Apps by Deploying ML Models as API Services
37mBuild a Data-Driven Web App That Everyone Can Use
37mSoftware Development for Machine Learning in Python
15mA Fiery approach to R web services
45mDjango for AI: Deploying Machine Learning Models with Django with Will Vincent
30mPython based predictive analytics with GraphLab Create
21mDeploying End-To-End Data Science with Shiny, Plumber, and Pins
01h 19mDatascience on the web