DP-100: Prepare Azure Model Deployment Competency (Intermediate Level)

Topics covered

• access diagnostic logs to monitor Data Lake Storage Gen2
• analyze model fairness using the Fairlearn Python package to identify imbalances between predictions and prediction performance
• collect and analyze Azure resource log data
• configure Azure Virtual Machines for disaster recovery
• configure, manage, and view activity log alerts using the Azure Monitor
• consume a real-time service that can be used to predict labels
• create an explainer and upload the explanation so it is available later analysis
• create Azure Data Factory linked services and datasets
• create Azure Data Factory pipelines and activities
• deploy a model as a real-time service to different compute targets
• describe cloud optimization, as well as best practices for optimizing data using data partitions, Azure Data Lake Storage tuning, Azure Synapse Analytics tuning, and Azure Databricks auto-optimizing
• describe data privacy problems and how differential privacy works
• describe high availability and disaster recovery and how they are related to SQL Server on Azure Virtual Machines
• describe how learning models can use global and local features to quantify the importance of each feature
• describe how model explainers can be created using the Azure Machine Learning SDK
• describe how to optimize Azure Cosmos DB using indexing and partitioning
• describe how training models can be biased due to biases in the training data
• describe methods for optimizing Azure Blob Storage
• describe the concepts of linked services and datasets and how they relate to the Azure Data Factory
• describe the concepts of pipelines and activities and how they relate to the Azure Data Factory
• describe the Integration Runtime and how it works with the Azure Data Factory
• describe the options for backing up, storing, and restoring SQL Server databases on virtual machine instances
• describe the purpose and features of Azure Always On availability groups
• describe the technical options for providing business continuity to SQL Server
• monitor a model that is deployed as an Azure Machine Learning real-time service using Jupyter Notebook and Python
• monitor Azure Blob storage
• monitor the Azure Cosmos DB using the portal and resource logs
• monitor the Azure Synapse Analytics jobs and the adaptive cache
• perform queries against the Azure Monitor logs
• publish and track Machine Learning pipelines and share them with others
• query Azure Log Analytics and filter, sort, and group query results
• schedule a Machine Learning pipeline based on an elapsed time or file system changes
• trigger a pipeline manually or using a schedule
• use a Jupyter Notebook and Python to detect and mitigate unfairness in a trained model
• use a Jupyter Notebook and Python to generate explanations that are part of a model training experiment
• use Azure Machine Learning pipelines to import, transform, and move data between steps
• use ML Studio to visualize data drift
• use the Azure Data Factory Analytics solution to monitor pipelines
• use the Azure Machine Learning SDK to create and run machine learning pipelines
• use visualizations in Azure Machine Learning Studio to visualize model explanations