Domain 2 of 5 · Chapter 3 of 3

Azure ML Service

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Microsoft Azure AI Fundamentals premium course — no separate purchase.

Included in this chapter:

  • What you get inside the workspace: the eight capabilities AI-900 names
  • Build-to-deploy on Azure ML: which capability serves which lifecycle stage
  • Exam-pattern recognition: AutoML vs Designer vs code, and the distractor traps

Authoring experiences in Azure Machine Learning studio

ExperienceCoding requiredWho it's forEffort / control trade-offBest when
Automated ML (AutoML)No-code (studio UI) or code (SDK/CLI)Analysts and developers without deep ML expertiseLowest effort; least control: the service picks algorithm and tunesYou want the best model found automatically by trying many algorithms
DesignerNo-code (drag-and-drop)Users who want a visual pipeline without writing codeLow effort; moderate control over the visual pipeline stepsYou want a visible, repeatable, editable training/inference pipeline
Notebooks + Python SDK / CLICode-firstData scientists and ML engineersHighest effort; full control over data, algorithm, and training loopYou need custom logic, frameworks, or fine-grained control

Decision tree

Which authoring approach in Azure Machine Learning studio? Want the service to try many models and pick the best? Automated ML (AutoML) no-code; auto-trains & ranks Yes — try many Want a visual drag-and-drop pipeline (no code)? No Designer visual pipeline; editable Yes Notebooks + Python SDK / CLI code-first; full control No — need code control
Cheat sheet

Cheat sheet

  • Azure ML is the platform for custom models
  • Workspace is the top-level container
  • Four ways to work with Azure ML
  • AutoML auto-trains and ranks many models
  • AutoML automates featurization and ensembling
  • Designer is no-code drag-and-drop pipelines
  • AutoML vs Designer: both no-code, decide on intent
  • Notebooks and SDK are the code-first path
  • Compute instance vs compute cluster
  • Datastores connect storage, data assets version data
  • Model registry plus MLflow for versioned, reproducible models
  • Managed online endpoint serves real-time scoring
  • Batch endpoint scores large datasets asynchronously
  • Responsible AI dashboard bundles fairness, interpretability, error analysis
  • Responsible AI dashboard needs registered MLflow scikit-learn models
  • Environments encapsulate software packages and runtime config
  • Azure ML data labeling: consensus and ML-assisted labeling
  • Azure ML workspace assets: experiments, components, storage, compute
  • Azure Custom Vision: training/prediction resources and compact domains

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. What is Azure Machine Learning?
  2. What is automated machine learning (AutoML)?
  3. What is Azure Machine Learning designer?
  4. What are compute targets in Azure Machine Learning?
  5. MLflow and Azure Machine Learning
  6. Endpoints for inference in Azure Machine Learning
  7. What is the Responsible AI dashboard?