One platform, four moving parts.
Most teams assemble this from five vendors and spend the year wiring them together. Sown Data ships them already connected.
Workspaces
Compute is a workspace, not a notebook. Choose a runtime version when you start one, then attach work to it. Runtimes bundle an Apache Spark™ version with libraries that are known to work together.
Warehouses
SQL warehouses run Trino™, with autoscaling bounded by a minimum, a maximum and a scaledown grace period. Warehouses are their own object with their own permissions and their own cost line.
Catalog
Apache Polaris holds the tables. Domains and environments give the catalog its shape, so finance_dev and finance_prod are separate namespaces with separate admins.
Sharing
No copy step. Publishing a table grants a scoped Polaris catalog role for that table alone, and the recipient — another internal domain or an external partner — reads it over the Iceberg REST catalog under vended, short-lived, table-scoped credentials.
Frequently asked.
Which table formats are supported?
Delta, Iceberg, Hudi and plain Parquet.
Do we have to move our data?
No. Storage stays where it is. The agent runs compute against it in place.
See it against your own data
We will walk through your workloads and show what runs where.