Thunderstorm forecasting with MLflow tracking

The scenario

StormCast teaches disciplined experimentation: time-aware train/test splits, feature lags, several model families, and MLflow tracking and model registry so you always know which model is in production and why.

System architecture

High-level flow of StormCast, from input to output.

What you will learn

• Time-series aware validation
• Experiment tracking
• Model registry and promotion

What you will build

• Feature pipeline
• Tracked experiments
• Forecast API

Build plan

1. Data
2. Features
3. Experiments
4. Registry
5. API

Tech stack