OCI Data Science Pro study materials
OCI Data Science Pro study resources

OCI Data Science Pro: Data Science Overview and OCI Data Science Concepts

OCI Data Science Pro · Work through your exam, one topic at a time.

All guide sections

Data Science Overview and OCI Data Science Concepts

Domain 1 - Data Science Overview and OCI Data Science Concepts

This domain tests whether you can place OCI Data Science in the OCI resource model and choose the right managed resource for each stage of the ML lifecycle. Expect questions on compartments, IAM, regional behavior, projects, notebook sessions, jobs, model catalog, deployments, and pipelines - and on what the service manages versus what you manage.

Core rules

  • OCI Data Science is a regional, managed service. Resources live in a compartment and are governed by IAM policies; a missing policy is a common cause of "cannot create/access" errors.
  • A project is a logical container for notebook sessions, jobs, models, and pipelines. It is not compute. Compute is attached to notebook sessions and job runs.
  • The service provides managed conda environments; you can also publish custom environments to Object Storage and import them.
  • The Accelerated Data Science (ADS) SDK is the Python surface for datasets, model evaluation, explainability, saving, and deploying models.
  • Separate exploration from production: notebooks for interactive work; jobs for repeatable batch training; model catalog for artifacts and provenance; model deployment for a hosted endpoint; pipelines for orchestration.

Worked example

A team must train a model weekly and serve it over HTTPS. Put the project in the team compartment and grant data scientists manage on data-science-family resources and use on the model deployment. Use a notebook session for exploration, a job with an output URI for training, register the artifact in the model catalog, then create a model deployment in the application subnet. The weekly cadence belongs in a schedule or pipeline, not a notebook.

Traps

  • Treating a project as a compute shape; projects do not run code.
  • Creating resources in the wrong region or compartment and then debugging IAM.
  • Assuming a notebook session is a production endpoint.
  • Forgetting that private subnets need a service gateway for Object Storage and a NAT gateway for internet.