Skip to main content

Challenge 09: Azure Machine Learning Workspace

Estimated Time

25-35 min | Cost: Free | Domain: Machine Learning on Azure (15-20%)

Exam skills covered

  • Describe capabilities of Automated Machine Learning
  • Describe data and compute services for machine learning
  • Describe model management and deployment capabilities in Azure ML

Overview

Azure Machine Learning (Azure ML) is a cloud platform for building, training, and deploying machine learning models. The workspace is the central hub — think of it as your ML laboratory that contains all your experiments, data, models, and compute resources in one organized space.

Think of Azure ML workspace like a professional kitchen. The kitchen (workspace) contains ingredients (datasets), cooking equipment (compute), recipes (notebooks/pipelines), finished dishes (trained models), and a serving counter (endpoints for deployment). Everything you need for the entire ML lifecycle is in one place.

Azure ML supports three main approaches: Automated ML (the platform builds models for you), Designer (drag-and-drop visual pipelines), and Notebooks (write code directly). For the AI-900 exam, focus on understanding what each component does rather than how to code with it.

Explore

Task 1: Azure ML workspace components

ComponentPurposeAnalogy
WorkspaceTop-level container for all ML assetsThe laboratory
DatastoresConnections to data storage (Blob, SQL, etc.)Ingredient pantry
DatasetsRegistered, versioned collections of dataPrepared recipe ingredients
Compute instancesVMs for development (notebooks)Your personal workstation
Compute clustersScalable VMs for training jobsIndustrial kitchen ovens
ExperimentsRecords of training runs with metricsLab notebook with results
ModelsTrained ML models (registered and versioned)Perfected recipes
EndpointsDeployed models serving predictionsRestaurant serving counter
PipelinesAutomated workflows (data prep → train → deploy)Assembly line

Task 2: Navigate Azure ML Studio

  1. Visit ml.azure.com (Azure Machine Learning Studio)
  2. If you don't have a workspace, explore the interface conceptually:
    • Left menu: Author (Notebooks, Automated ML, Designer) | Assets (Data, Jobs, Models, Endpoints) | Manage (Compute)
  3. Key areas to understand for the exam:
    • Automated ML: Upload data → choose target column → Azure builds the best model
    • Designer: Visual drag-and-drop pipeline builder
    • Models: Registry of all trained models with versions
    • Endpoints: Where deployed models serve predictions

Task 3: Understand Automated ML capabilities

Automated ML (AutoML) automates the most time-consuming parts of machine learning:

StepWhat AutoML does automatically
Algorithm selectionTests multiple algorithms (logistic regression, decision trees, gradient boosting, etc.)
Hyperparameter tuningFinds the best settings for each algorithm
Feature engineeringCreates and selects the most useful features
Model evaluationCompares all models using appropriate metrics
Best model selectionReturns the top-performing model

AutoML supports three task types:

  • Classification — predict categories
  • Regression — predict numbers
  • Time-series forecasting — predict future values over time

Task 4: Model deployment and endpoints

Once a model is trained, it needs to be deployed so applications can use it:

Deployment typeUse caseHow it works
Real-time endpointImmediate predictionsApplication sends data, gets prediction back instantly (REST API)
Batch endpointProcess large datasetsSubmit a dataset, get predictions back later (asynchronous)

Deployment workflow:

  1. Register the model in the Model Registry
  2. Create an endpoint (real-time or batch)
  3. Deploy the model to the endpoint
  4. Test by sending data and receiving predictions
  5. Monitor performance and data drift over time
Azure CLI Alternative
# List Azure ML workspaces in your subscription
az ml workspace list --output table

# List compute instances in a workspace
az ml compute list --workspace-name my-workspace --resource-group my-rg --output table

# List registered models
az ml model list --workspace-name my-workspace --resource-group my-rg --output table

Key Concepts

ConceptDefinition
Azure ML workspaceCentral hub containing all ML resources: data, compute, models, endpoints
Automated ML (AutoML)Feature that automatically trains and compares multiple models, selecting the best one
Compute instanceSingle VM for development and testing (notebooks)
Compute clusterScalable group of VMs that grows/shrinks based on training job demand
Model registryVersioned catalog of trained models for tracking and deployment
Real-time endpointDeployed model that returns predictions immediately via REST API
Batch endpointDeployed model that processes large datasets asynchronously
ML pipelineAutomated, repeatable workflow for the full ML process
DesignerVisual drag-and-drop tool for building ML pipelines without code

Common Misconceptions

MisconceptionReality
"You need to know how to code to use Azure ML"Azure ML Designer provides a drag-and-drop visual experience, and Automated ML builds models with minimal configuration. Code (Python) is optional
"Automated ML produces production-ready models every time"Automated ML gives you a great starting point, but production deployment often requires additional evaluation, testing, and monitoring
"A compute instance is required to deploy a model"Compute instances are for development. Deployed models run on managed endpoints — separate infrastructure optimized for serving predictions
"Once deployed, a model never needs updating"Models degrade over time as real-world data changes (concept drift / data drift). Monitoring and retraining are ongoing needs
"Azure ML and Azure AI services are the same thing"Azure AI services provide pre-built, ready-to-use AI APIs. Azure ML is for building your OWN custom models from your data

Knowledge Check

1. A data scientist wants Azure to automatically try multiple algorithms and select the best-performing model for their dataset. Which Azure ML feature should they use?

2. What is the purpose of a real-time endpoint in Azure Machine Learning?

3. Which Azure ML component provides a visual, drag-and-drop experience for building machine learning pipelines?

4. What is the difference between a compute instance and a compute cluster in Azure ML?

5. What is the primary difference between Azure Machine Learning and Azure AI services?

Learn More