Since Azure Databricks was first introduced, it has exposed Spark machine learning (ML) features. But more recently, the plot has thickened, with the addition of Databricks-specific ML features. These include machine learning operations (MLOps) capabilities via both a hosted version of the open source MLflow project and a new proprietary feature store; and, to make machine learning accessible to all developers, a new automated machine learning (AutoML) engine.
In this session we’ll walk through these newer capabilities and look at an end-to-end demo, where we’ll build an ML model, manage its features, deploy it to production and generate predictions from it using both interactive and batch scoring.
You will learn:
- The basics of automated machine learning (AutoML) and the specifics of AutoML in Databricks
- The use of MLflow to manage experiments, track models in the repository and deploy them
- What feature stores are and how to use Databricks’ implementation
- How to view and edit AutoML-generated notebook code