Amazon SageMaker Unified Studio
The data and AI workspace of the next-generation Amazon SageMaker. One catalog for data on AWS, Snowflake and Databricks; domains and projects so each team owns and shares its data; blueprints so the platform team hands out governed environments; and a Data Agent to query it all in plain language.
Domains own projects, the platform team runs their workspaces, the catalog moves data between them.
- 1The platform team turns blueprints into project profiles: which tools, which account, which settings are locked.
- 2Each domain unit creates its projects from a profile; every project gets its own workspace.
- 3Sources are registered or federated in the lakehouse catalog and reach the producer workspace.
- 4The Sales team builds a data product in its workspace.
- 5It publishes the product to the catalog with business names, glossary terms and lineage.
- 6The Marketing project finds it and requests a subscription; the Sales owner approves.
- 7Access is granted to the Marketing project only, in its own workspace.
- 8Analysts query it there, in SQL or through the Data Agent in plain language.
Snowflake and other databases connect through a Glue connection and are federated into the catalog without copying data. Databricks is federated through Glue Data Catalog federation to Unity Catalog: Iceberg tables directly, Delta tables once UniForm writes their Iceberg metadata. Lineage accepts OpenLineage events from any system that emits them.
One catalog, team ownership, governed workspaces, and an agent.
Catalog beyond AWS
S3, Redshift and databases sit next to federated Snowflake and Databricks tables. Each asset gets business names, glossary terms, metadata forms and lineage, so people search by meaning, not by table name.
Domains own their data
Domain units mirror your organisation and carry its policies. A producer project publishes data products; a consumer project subscribes, the owner approves, and access is granted for that project only.
Environments on demand
Admins define project profiles from blueprints: which tools a project gets, in which account, with which settings locked. Custom blueprints are your own CloudFormation templates, versioned like any other code.
Chat with your data
The SageMaker Data Agent works in notebooks and the query editor. It reads the catalog, including business context, to write SQL and Python from a plain-language request, and keeps the conversation going for follow-ups.
Three decisions shape how far it scales.
How far the catalog reaches
Federation brings Snowflake and Databricks tables into the same search, glossary and lineage as your AWS data. The data stays where it is; the catalog holds the metadata and the access path.
How you cut domains and projects
Cut domain units along data ownership, which changes less often than reporting lines. Projects hold a team's inventory until it publishes; consumers subscribe to published assets and the owner approves each request.
Built-in or custom blueprints
Start from built-in profiles, then assemble your own from built-in and custom blueprints. Each custom blueprint has a provisioning role and the domain units allowed to use it; only the parameters you leave editable are open to project owners.
Tag values describe mechanisms, not measured figures.
A Unified Studio bill: the catalog, then the services you use.
Build & optimize your data platform on AWS
State of the art, costs under control. Most bills grow with habits, not with data.
Read the approach →The one-page Amazon SageMaker Unified Studio field guide.
Best practices, cost traps and performance checks, in the order you should run them. Printable, no fluff.