Databricks Platform Administrator

Manual user management does not survive a growing company. You bring in users from an identity provider, keep groups in sync, and use APIs and the command line to manage workspaces without clicking through each setting.

• Single sign-on and the identity provider
• Automatic provisioning of users and groups
• Tokens and service principals for automation
• The REST API, SDKs and the CLI for administration
• Auditing changes made by automation

You build A provisioning setup plus a script that audits group membership through the API.

Compute is where money is spent and where problems start. You set up clusters with the right node types and init scripts, write policies that stop expensive mistakes, and use pools to cut start-up time.

• Cluster modes and access modes
• Node types and runtime versions
• Init scripts and libraries
• Cluster policies as guardrails
• Pools and start-up time

You build A set of cluster policies for three kinds of users, with a pool for interactive work.

Not all work needs a cluster. You configure SQL warehouses for analytics, decide where serverless compute fits, and understand how Model Serving endpoints are provisioned and who may use them.

• SQL warehouse types and sizes
• Scaling, auto-stop and queuing
• Serverless compute: benefits and limits
• Model Serving endpoints and their permissions
• Choosing compute for an analytics team

You build A warehouse and serving configuration chosen for a stated team and budget.

Production work runs on a schedule and under a service identity. You schedule jobs and pipelines, choose the compute and permissions they use, and set up the notifications that tell an owner when something fails.

• Jobs, tasks and schedules
• Pipelines and their compute
• Run-as identities and permissions
• Notifications and failure handling
• Controlling who can run, edit and view

You build A scheduled job and pipeline running as a service principal with alerts to its owner.

Access works best when it is defined once and inherited. You set permissions on workspace items and data, compare managed and external tables and views, and choose an access model under Unity Catalog.

• Permissions on notebooks, folders and compute
• Securable data objects and inheritance
• Managed and external tables
• Views for restricted access
• Choosing an access model for a team

You build A permission model for one catalog with views for a restricted audience.

Sensitive values and shared data need extra care. You store credentials as secrets, share data safely with Delta Sharing, and control who can create, view and use machine learning models.

• Secret scopes and who can read them
• Keeping credentials out of notebooks
• Delta Sharing: providers, recipients and limits
• Permissions on registered models
• Reviewing the security of a workspace

You build A secrets setup, a limited data share and a model permission plan.

You set up a workspace for a new team from nothing: identity, compute, scheduled work, data access and secrets. Then you check what the team can and cannot do and fix what surprises you.

• Gathering the team’s needs
• Setting up identity and groups
• Provisioning compute with policies
• Granting data access and handling secrets
• Testing the setup as a team member

You build A documented workspace configuration for a new team, tested from a user’s point of view.