Design ML Solutions (30-35%) Intermediate

This is the highest-weighted domain on the DP-100 exam. You must understand how to design the Azure ML workspace, choose compute targets, manage data assets, and build pipeline architectures. Master this domain and you are a third of the way to passing.

Azure Machine Learning Workspace

The workspace is the top-level resource for Azure ML. Everything — compute, data, models, endpoints, experiments — lives inside a workspace.

Key Workspace Components

# Creating a workspace with Python SDK v2
from azure.ai.ml import MLClient
from azure.ai.ml.entities import Workspace
from azure.identity import DefaultAzureCredential

# Authenticate
credential = DefaultAzureCredential()

# Define workspace
ws = Workspace(
    name="dp100-exam-workspace",
    location="eastus",
    display_name="DP-100 Study Workspace",
    description="Workspace for DP-100 exam preparation",
    tags={"purpose": "certification-prep"}
)

# Create workspace (also creates Storage, Key Vault, App Insights)
ml_client = MLClient(
    credential=credential,
    subscription_id="your-subscription-id",
    resource_group_name="dp100-rg"
)
ml_client.workspaces.begin_create_or_update(ws)

Compute Targets

Choosing the right compute is critical for both the exam and real projects. Know when to use each type.

# Creating a compute cluster with SDK v2
from azure.ai.ml.entities import AmlCompute

# Define compute cluster
cluster = AmlCompute(
    name="dp100-cluster",
    type="amlcompute",
    size="Standard_DS3_v2",      # VM size
    min_instances=0,              # Scale to zero when idle
    max_instances=4,              # Max parallel nodes
    idle_time_before_scale_down=120,  # Seconds before scale-down
    tier="Dedicated"              # or "LowPriority" for spot VMs
)

ml_client.compute.begin_create_or_update(cluster)

Data Assets and Datastores

Azure ML uses datastores to connect to storage services and data assets to reference specific datasets.

Datastore Types

• Azure Blob Storage — Default datastore, best for unstructured data (images, text files)
• Azure Data Lake Storage Gen2 — Hierarchical namespace, best for large-scale analytics
• Azure SQL Database — Structured data, direct SQL queries
• Azure Files — File shares, mountable as network drive

Data Asset Types

# Register a data asset (URI File)
from azure.ai.ml.entities import Data
from azure.ai.ml.constants import AssetTypes

# CSV file in blob storage
data_asset = Data(
    name="customer-churn-data",
    description="Customer churn dataset for training",
    path="azureml://datastores/workspaceblobstore/paths/data/churn.csv",
    type=AssetTypes.URI_FILE,
    version="1"
)

ml_client.data.create_or_update(data_asset)

# Register an MLTable asset
mltable_asset = Data(
    name="customer-churn-mltable",
    description="Customer churn as MLTable with schema",
    path="azureml://datastores/workspaceblobstore/paths/data/churn-mltable/",
    type=AssetTypes.MLTABLE,
    version="1"
)

ml_client.data.create_or_update(mltable_asset)

Environments

Environments define the software dependencies for training and deployment. Azure ML supports curated environments (pre-built) and custom environments.

# Using a curated environment
from azure.ai.ml.entities import Environment

# List curated environments
envs = ml_client.environments.list()
for env in envs:
    if "sklearn" in env.name.lower():
        print(f"{env.name}: {env.version}")

# Create custom environment from conda file
custom_env = Environment(
    name="dp100-custom-env",
    description="Custom environment for DP-100 training",
    conda_file="conda.yml",
    image="mcr.microsoft.com/azureml/openmpi4.1.0-ubuntu20.04:latest",
    version="1"
)

ml_client.environments.create_or_update(custom_env)

Pipeline Architecture

Pipelines orchestrate multi-step ML workflows. For the exam, understand pipeline components, data passing, and scheduling.

# Building a pipeline with SDK v2
from azure.ai.ml import dsl, Input, Output
from azure.ai.ml.entities import CommandComponent

# Define pipeline steps as components
@dsl.pipeline(
    description="DP-100 training pipeline",
    compute="dp100-cluster"
)
def training_pipeline(input_data):
    # Step 1: Data preparation
    prep_step = prep_component(
        raw_data=input_data
    )

    # Step 2: Train model
    train_step = train_component(
        training_data=prep_step.outputs.prepared_data,
        learning_rate=0.01,
        n_estimators=100
    )

    # Step 3: Evaluate model
    eval_step = eval_component(
        model=train_step.outputs.model,
        test_data=prep_step.outputs.test_data
    )

    return {
        "trained_model": train_step.outputs.model,
        "metrics": eval_step.outputs.metrics
    }

# Submit pipeline
pipeline_job = training_pipeline(
    input_data=Input(
        type="uri_file",
        path="azureml://datastores/workspaceblobstore/paths/data/churn.csv"
    )
)

ml_client.jobs.create_or_update(pipeline_job)

Practice Questions