MLOps on Databricks: Building a Model Factory with MLflow + Kedro

We designed a model factory to automate retraining and deployment. MLflow manages experiments and registry; Kedro standardizes data/feature pipelines. This reduced ongoing dependency on data scientists and improved time-to-production.

Core components

• Versioned feature store and model registry integration
• Automated retraining triggers (data or performance drift)
• CI/CD with staged deployments and rollbacks

These patterns generalize beyond mining to any domain needing reliable, repeatable ML at scale.