As AI development teams grow, managing who can access which computing resources, models, and jobs across multiple tools becomes a coordination problem. A workspace in PAI centralizes resource access, member permissions, and AI asset management under a single organizational unit—so teams can collaborate across development stages without fragmented permission setups.
What workspaces provide
Workspaces offer the following capabilities:
Computing resource management: Manage all AI development computing resources from a single platform.
Permission management: Control what each team member can do at a fine-grained level.
AI asset management: Save and reuse models and algorithms across training runs and team members.
Event monitoring and alerting: Set up alert rules to track Deep Learning Containers (DLC) jobs, Machine Learning Designer pipeline jobs, or model version approval status.
Resource scheduling: Allocate resource quotas, cap job runtimes, and assign job priorities by component and role to distribute resources efficiently.
Roles
PAI provides predefined roles that map to common responsibilities in AI development teams. Assign roles based on what each member needs to do in the workspace.
| Role | Description |
|---|---|
| Resource administrator | Must be assigned to an Alibaba Cloud account or authorized RAM user. Adds or deletes workspaces, and creates Data Science Workshop (DSW) instances. |
| Workspace administrator/owner | Manages members and public assets in the workspace. |
| Algorithm developer | Manages training jobs and public assets. |
| Algorithm O&M engineer | Manages all training jobs in the workspace, including viewing DLC jobs created by other members. |
| Labeling administrator | Creates and updates labeling datasets. |
| Visitor | Views workspace information: members, resource groups, and pipelines. |
| MaxCompute developer | Uses MaxCompute to execute jobs submitted in the workspace. |
For the full permissions matrix for each role, see Roles and permissions.
Next steps
To create and manage workspaces or configure notifications, see Workspace.
To manage AI assets (datasets, models, images, and jobs) in PAI, see AI Computing Asset Management.