Designing and Implementing a Data Science Solution on Azure

 

• Candidates for this exam should have subject matter expertise applying data science and machine learning to implement and run machine learning workloads on Microsoft Azure.
Responsibilities for this role include planning and creating a suitable working environment for data science workloads on Azure. They run data experiments and train predictive models. In addition, they manage, optimize, and deploy machine learning models into production.
• A candidate for this certification should have knowledge and experience in data science and using Azure Machine Learning and Azure Databricks.

Manage Azure resources for machine learning (25–30%) Create an Azure Machine Learning workspace

• Create an Azure Machine Learning workspace
• Configure workspace settings
• Manage a workspace by using Azure Machine Learning studio

Manage data in an Azure Machine Learning workspace

• Select Azure storage resources
• Register and maintain datastores
• Create and manage dataset

Manage compute for experiments in Azure Machine Learning

• Determine the appropriate compute specifications for a training workload
• Create compute targets for experiments and training
• Configure Attached Compute resources including Azure Databricks
• Monitor compute utilization

Implement security and access control in Azure Machine Learning

• Determine access requirements and map requirements to built-in roles
• Create custom roles
• Manage role membership
• Manage credentials by using Azure Key Vault

Set up an Azure Machine Learning development environment

• Create compute instances
• Share compute instances
• Access Azure Machine Learning workspaces from other development environments

Set up an Azure Databricks workspace

• Create an Azure Databricks workspace
• Create an Azure Databricks cluster
• Create and run notebooks in Azure Databricks
• Link and Azure Databricks workspace to an Azure Machine Learning workspace

Run experiments and train models (20–25%) Create models by using the Azure Machine Learning designer

• Create a training pipeline by using Azure Machine Learning designer
• Ingest data in a designer pipeline
• Use designer modules to define a pipeline data flow
• Use custom code modules in designer

Run model training scripts

• Create and run an experiment by using the Azure Machine Learning SDK
• Configure run settings for a script
• Consume data from a dataset in an experiment by using the Azure Machine Learning SDK
• Run a training script on Azure Databricks compute
• Run code to train a model in an Azure Databricks notebook

Generate metrics from an experiment run

• Log metrics from an experiment run
• Retrieve and view experiment outputs
• Use logs to troubleshoot experiment run errors
• Use MLflow to track experiments
• Track experiments running in Azure Databricks

Use Automated Machine Learning to create optimal models

• Use the Automated ML interface in Azure Machine Learning studio
• Use Automated ML from the Azure Machine Learning SDK
• Select pre-processing options
• Select the algorithms to be searched
• Define a primary metric
• Get data for an Automated ML run
• Retrieve the best model

Tune hyperparameters with Azure Machine Learning

• Select a sampling method
• Define the search space
• Define the primary metric
• Define early termination options
• Find the model that has optimal hyperparameter values

Deploy and operationalize machine learning solutions (35–40%) Select compute for model deployment

• Consider security for deployed services
• Evaluate compute options for deployment

Deploy a model as a service

• Configure deployment settings
• Deploy a registered model
• Deploy a model trained in Azure Databricks to an Azure Machine Learning endpoint
• Consume a deployed service
• Troubleshoot deployment container issues

Manage models in Azure Machine Learning

• Register a trained model
• Monitor model usage
• Monitor data drift

Create an Azure Machine Learning pipeline for batch inferencing

• Configure a ParallelRunStep
• Configure compute for a batch inferencing pipeline
• Publish a batch inferencing pipeline
• Run a batch inferencing pipeline and obtain outputs
• Obtain outputs from a ParallelRunStep

Publish an Azure Machine Learning designer pipeline as a web service

• Create a target compute resource
• Configure an inference pipeline
• Consume a deployed endpoint

Implement pipelines by using the Azure Machine Learning SDK

• Create a pipeline
• Pass data between steps in a pipeline
• Run a pipeline
• Monitor pipeline runs

Apply ML Ops practices

• Trigger an Azure Machine Learning pipeline from Azure DevOps
• Automate model retraining based on new data additions or data changes
• Refactor notebooks into scripts
• Implement source control for scripts

Implement responsible machine learning (5–10%) Use model explainers to interpret models

• Select a model interpreter
• Generate feature importance data

Describe fairness considerations for models

• Evaluate model fairness based on prediction disparity
• Mitigate model unfairness

Describe privacy considerations for data

• Describe principles of differential privacy
• Specify acceptable levels of noise in data and the effects on privacy