MLOps by example: seven lessons and a Danish energy forecast

What it does

This repository is a curriculum, not a library. It walks through building an end-to-end batch ML system that forecasts hourly electricity consumption across Denmark. The codebase is split into a pipeline trio—feature engineering, training, and batch prediction—and a web app trio for serving, visualization, and monitoring, all orchestrated by Airflow and containerized with Docker.

The interesting bit

Instead of inventing a new framework, it serves as opinionated glue code that wires together a standard MLOps toolkit: Hopsworks, Weights & Biases, Great Expectations, FastAPI, Streamlit, and GCP. The refreshing part is a bonus lesson that openly dissects the architecture’s imperfections and trade-offs rather than pretending the design is flawless.

Key highlights

• Uses a real Danish energy dataset with hourly consumption readings by price area and consumer type, though the original API is now obsolete and replaced by a static file.
• Integrates a feature store, experiment tracker, model registry, data validation, and CI/CD into one coherent batch architecture.
• Bundles 2.5 hours of reading and video material hosted on Medium, though the narrative walkthrough sits behind a paywall.
• Provides explicit cost guidance: expect roughly $20 in GCP fees to run the final deployment.
• Targets intermediate-to-advanced MLEs and software engineers transitioning into ML roles.

Caveats

• The upstream data API is obsolete; the course now relies on a mocked static file downloaded from Google Drive, which limits real-time experimentation.
• The setup is tested only on Ubuntu 20.04/22.04 with Python 3.9, so mileage on other platforms is unclear.
• Several lessons and the detailed walkthrough live behind Medium’s paywall, so the “free” label applies mainly to the source code.

Verdict

Grab this if you want a guided, opinionated tour of a full MLOps stack and don’t mind paying for Medium and GCP to complete the arc. Skip it if you are looking for a reusable library or a tool-agnostic framework.