Azure RAG Patterns: Azure AI Search & OpenAI Pipelines

Course Overview

Master Retrieval Augmented Generation (RAG) on Azure by learning to build production-ready AI applications using Azure AI Search, Azure OpenAI, hybrid retrieval, and agentic chat. This hands-on course guides you through secure document ingestion, hybrid retrieval, cited chat, evaluation, and deployment using Microsoft Foundry and Azure services.

Key Takeaways

• Build automated Python ingestion pipelines to parse, chunk, and index documents with Hybrid Search and Semantic Ranker.
• Design production RAG topologies using Azure AI Foundry SDK, query rewriting, and conversation state management.
• Deploy safe RAG applications to Azure App Service with Azure Content Safety guardrails and automated RAG Triad evaluations.
• Implement end-to-end security, Azure AI Content Safety guardrails, and automated evaluations for production deployment.
• Prepare, chunk, enrich, embed, and index enterprise documents using various retrieval patterns including keyword, vector, hybrid, semantic, and agentic retrieval.
• Build chat orchestration flows with bounded context and source citations.
• Test for prompt injection and apply retrieval-time security filters to minimize sensitive data exposure.
• Evaluate retrieval relevance, groundedness, citation coverage, latency, and cost with a production-readiness checklist.

Prerequisites

• Basic proficiency in Python.
• Active Azure subscription with access to Azure OpenAI services.
• Basic familiarity with Azure resources and REST APIs.

Target Learners

• Azure developers building enterprise AI applications.
• AI engineers and data engineers responsible for search, indexing, or model integration.
• Solution architects designing secure RAG and chat systems.
• Technical leads needing evaluation and production-readiness approaches.
• Developers with basic Python or REST API knowledge seeking end-to-end Azure implementation.

Final Project

Build a cited internal policy assistant application that ingests controlled documents, stores structured chunks in Azure AI Search, retrieves evidence with hybrid search, and sends bounded context to a deployed model via Microsoft Foundry workflows. The app returns answers with source references. You will run tests including golden-question sets, out-of-scope questions, and prompt-injection documents, record quality and latency results, and complete a production-readiness checklist.