Building AI Agents with Langchain and Microsoft Azure

Course Overview

This hands-on course teaches how to build modern AI agents from scratch using LangChain and Microsoft Azure. It emphasizes practical implementation over theory, guiding learners through deploying models in Azure AI Foundry, building ReAct agents, and extending capabilities with tools, function calling, and integration with external systems like MCP servers. The course also covers sandboxed execution environments, memory management with Azure Cosmos DB, human-in-the-loop workflows, middleware hooks, observability with LangSmith, and multi-agent orchestration.

Key Takeaways

• Build AI agents using LangChain and LangGraph with Azure AI Foundry.
• Deploy and configure large language models on Azure using Terraform and secure API access.
• Design agents with tools, function calling, and the ReAct reasoning loop.
• Connect agents to data using MCP (Microsoft Learn + web search).
• Execute Python and shell code safely using sandboxed environments.
• Add memory to agents using Azure Cosmos DB for multi-turn conversations.
• Implement human-in-the-loop workflows and model guardrails.
• Observe, debug, and optimize agents using LangSmith tracing.
• Design multi-agent orchestration systems.

Prerequisites

• Basic knowledge of Python (functions, APIs, packages).
• Familiarity with REST APIs and JSON is helpful.
• Azure account (free tier works) to deploy models and services.
• Basic understanding of AI/LLMs concepts is a plus but not required.
• Development environment such as VS Code or Jupyter Notebook.
• Curiosity to build real-world AI agents.

Target Learners

• Python developers who want to build real AI agents.
• Azure engineers and cloud architects exploring AI workloads.
• AI/ML practitioners looking to move from prompts to autonomous agents.
• Developers interested in LangChain, MCP, and agent orchestration.
• Anyone curious about building production-ready AI systems on Azure.