Secure Azure LLM Application Architecture

This unit covers the design, assembly, and deployment of a three-tier LLM application in Microsoft Azure, including retrieval-augmented generation (RAG) techniques and the Azure services that support them. Students will work with Azure Front Door, App Service, API Management, Storage Accounts, and Cosmos DB while using references and documentation to inform secure, scalable architecture choices that prepare them for the capstone integration.

Learning Objectives

• Analyze the three-tier LLM application architecture in Azure by mapping components (Front Door, App Service, API Management, Storage Account, Cosmos DB), data flows, and security boundaries.
• Apply retrieval-augmented generation (RAG) principles by designing and implementing a retrieval pipeline that uses Azure Storage Account and/or Cosmos DB for document storage, embedding lookup, and prompt construction.
• Demonstrate configuring and deploying Azure infrastructure by provisioning Azure Front Door for global routing and WAF, deploying an Azure App Service for the LLM API, and implementing API Management policies for authentication, rate limiting, and logging.
• Evaluate storage and API management trade-offs (Blob Storage vs. Cosmos DB; caching, consistency, latency, and cost) and justify secure, scalable design decisions for a production-ready LLM application using authoritative references.