AI & Machine Learning on Azure

Overview of the Azure AI platform and core concepts.

• The Azure AI Platform (Azure ML, Cognitive Services, OpenAI).
• Review of ML Concepts and Model Lifecycles.
• Azure Role-Based Access Control (RBAC) for AI resources.
• Introduction to Azure Machine Learning Studio.

Preparing and managing data for machine learning in Azure.

• Using Azure Data Lake Storage (ADLS) Gen2.
• Azure Data Factory (ADF) for data ingestion pipelines.
• Data transformation with Azure Synapse Analytics or Databricks.
• Azure ML Datasets and Datastores.

Using the integrated workspace for machine learning.

• Navigating the Azure ML Studio (Notebooks, Designer, AutoML).
• Azure ML SDK v2 and CLI v2 for programmatic access.
• Automated ML (AutoML) for rapid model creation.
• Azure ML Designer for no-code/low-code pipelines.

Training custom models and managing compute resources.

• Configuring Compute Instances and Compute Clusters.
• Training custom models using script-based jobs (SDK/CLI).
• Distributed training on multiple nodes.
• Tracking experiments and logging metrics.

Optimizing and deploying models for real-time inference.

• Hyperparameter tuning with Hyperdrive (Sweep Jobs).
• Registering and versioning models in the Model Registry.
• Deploying models to Managed Online Endpoints (Real-time).
• Batch Endpoints for offline predictions.

Using pre-built vision services for common tasks.

• Azure Cognitive Services for Vision (Image Analysis, OCR).
• Azure Custom Vision for specialized image classification.
• Face API for detection and recognition.
• Azure Form Recognizer for document intelligence.

Using pre-built language and speech services.

• Azure Cognitive Service for Language (Sentiment, Key Phrases).
• Language Understanding (LUIS) for conversational AI.
• Speech Service (Speech-to-Text, Text-to-Speech).
• Translator Service.

Building enterprise applications with foundation models.

• Introduction to Azure OpenAI Service (GPT-4, DALL-E).
• Prompt Engineering and managing deployments.
• Building RAG applications with Azure AI Search (formerly Cognitive Search).
• Orchestrating workflows with Prompt Flow.

Automating the ML lifecycle from development to production.

• Azure ML Pipelines for CI/CD automation.
• Integrating with Azure DevOps or GitHub Actions.
• Model monitoring for data drift and performance.
• Managing ML environments and components.

Building a complete project from data to production endpoint on Azure.

• Define project scope (e.g., custom vision or language task).
• Build the data processing pipeline with ADF.
• Train, tune, and register the model with Azure ML Studio.
• Deploy the model to a Managed Online Endpoint and build a client.