modelsIntermediate

Machine Learning

Track experiments with MLflow, build features, register models in Unity Catalog and serve them behind an endpoint.

0 of 17 done
17 of 17 written|~50 h
Steps through the path one concept at a time, in order.100%
01Foundations
02Experiment
03Register and serve
01Architecture of the Data Intelligence Platform
02Columns, rows, and DataFrame structure
03MLflow tracking on Databricks
04Feature engineering and the feature store
05MLflow Tracing for GenAI applications
06AutoML
07Training and tuning a classic model
08MLflow 3 for models
09AI Runtime and serverless GPU compute
10Training sets and point-in-time joins
11Feature Views
12Models in Unity Catalog
13Model serving endpoints
14Serving compute and scaling
15Online Feature Store
16MLflow deployment jobs
17Monitoring a deployed model
FoundationsPlatform & governanceAI & ML
Drag to pan · scroll to zoom · click a task to open it · the small button says what it covers
The same path as a list17 concepts
Stage 1 · Foundations
Architecture of the Data Intelligence PlatformColumns, rows, and DataFrame structure
Stage 2 · Experiment
MLflow tracking on DatabricksFeature engineering and the feature storeMLflow Tracing for GenAI applicationsAutoMLTraining and tuning a classic modelMLflow 3 for modelsAI Runtime and serverless GPU computeTraining sets and point-in-time joinsFeature Views
Stage 3 · Register and serve
Models in Unity CatalogModel serving endpointsServing compute and scalingOnline Feature StoreMLflow deployment jobsMonitoring a deployed model

Everything here assumes the data is already governed and clean, so do Lakehouse Foundations first if that is not true yet.

The order matters: tracking gives you something to compare, features give you something reproducible, the registry gives you something to promote, and serving gives you something to call. Skipping to serving is how models end up unversioned in a notebook.

Full resources for this path

593 free
Newsletter

What changed, and what it means for you

Databricks renames and deprecates things often. Get the ones that actually change how you work, plus the concepts rewritten because of them. No more than the schedule below, and one click to leave.

Choose what you get5 of 43
Your stack · Platform & governance

Tick what you run, and untick “Everything worth knowing” to get only that.

Data engineering · Lakeflow
Data warehousing & BI
AI & ML
Your learning paths

Each one is a section of the letter, and you only get the ones you tick. Every email carries a link back to this list, so you can change it later without an account.

Double opt-in: we send one email to check the address is yours. Nothing else until you confirm.