SageMaker Unified Studio Governance
Documentation reviewed: 2026-09-12. Experiment outcomes refer to the 2026-09-01 record and the historical September 2 documentation.
Amazon SageMaker Unified Studio manages collaboration, tools and catalog assets for data/AI teams. This section explains domain/project boundaries for EKS pipeline assets, users and execution permissions. The Unified Studio/DataZone project here is not the same API object as a SageMaker AI MLOps Project or SageMaker AI Studio domain.
Boundaries covered
| Topic | What to verify |
|---|---|
| Domain type | IAM-based versus IAM Identity Center-based login/administration |
| Project profile / blueprint | Tools provisioned at creation versus enabled on demand |
| Member / execution role | Portal/project access identity versus AWS resource execution identity |
| Membership / data access | Administrative designations versus IAM, Lake Formation and catalog data permissions |
| Lifecycle | Project existence, environment readiness, actual tool access and owned-resource cleanup |
Part 4: Domain, project and membership explains the Qwen experiment's recorded failures through these boundaries. A Unified Studio project is this guide's governance choice, not a mandatory technical dependency for every SageMaker Training Job or EKS training workload.
Distinguish historical evidence from current state
The stored 2026-09-01 validation JSON records a stop before training, cleanup of experiment App/S3/IAM resources and one remaining Unified Studio project. September 2 documentation records an ACTIVE recheck at that time. This documentation review did not query the AWS account again and does not assert that one project still remains today.
Before resuming that experiment, an authorized operator must verify current inventory, membership and cleanup state. Do not generalize the old failure into an automatic grant of new privileges or deletion of shared resources.
Related guides: