README.md

Azure ML E2E Workshop (Updated to 2020)

Target Audience

Anyone who wants a comprehensive end-to-end understanding of Azure Machine Learning (AzureML).

Key Goals

1. Understand the core concepts of AzureML
2. Understand how to use AzureML in an end-to-end fashion
3. Serve as a reference for common scenarios

Agenda

1 - Workspace Concepts: infra setup, ARM, workspace setup, computes, datastores, setup

1. Set up your workspace and compute
2. Register a dataset
3. Run AutoML from the UI (optional)
4. Designer interface (optional)
5. Clone git repo to Compute Instance

2 - Datasets, Model Training (AML, HyperDrive and AutoML), Model Inference

AML training including HyperDrive:

1. Notebook for plain vanilla Scikit-Learn model training in AML local compute (AML VM)
2. Notebook for Scikit-Learn model training in AML remote compute and HyperDrive

Interpretability:

1. Notebook for Model Interpretability in AML on local compute (includes upload explaination to the experiment)
2. Notebook for Model Interpretability on AML Compute

Automated ML:

1. Notebook for AutoML local compute
2. Notebook for AutoML remote compute

Pipelines & Batch Inference:

1. Use a model for batch inference

3 - MLOps (model management, deployment, inference, automation)

1. Deploy a model to ACI with Python - Start with this!
2. Automate training & deployment - Basic example to get started with MLOps in Azure DevOps
3. Automate training & deployment using ML Pipelines Work in Progress!
4. Create automation workflow with EventGrid

4 - Enterprise Readiness

1. Enterprise security - VNet for AML Compute, Compute Instance, Key Vault, Storage, including secure deployment to AKS

Reading materials:

• Enteprise Security Overview
• Manage access via RBAC
• Managing compute quotas
• VNET integration
• Azure Monitor

5 - R Support

1. R Integration - Training with R, deployment with R, running a Shiny app