AI-103: Azure AI App and Agent Developer - Complete Course

Explore vector searching and grounding LLMs with embedded data in Foundry. Deploy text embedding models, index files, and connect knowledge bases to agents for scalable retrieval.

Demonstrates multi-agent orchestration by routing user prompts to simple or complex agents via visual workflows, deploying GPT-5 nano, and exposing endpoints with YAML and code.

Build and deploy a three-agent workflow for multi-agent orchestration, routing simple versus complex queries between a small and a large language model using code-driven deployment and streaming outputs.

Deploy an Azure AI agent and attach three MCP servers as tools, configure server labels and URLs in config.yaml and launch.json, and review hands-on traces of weather and campsite queries.

Set up tracing in Foundry by connecting AI insights resources and logging with log analytics and application insights, then explore tool calls, reasoning, and MCP server interactions to debug conversations.

Resolve a tracing setup bug in lab 9 by registering the operational insights resource provider in your azure subscription, then return to Foundry to create the new tracing resource.

Learn how to use agent memory in Azure Foundry, from short-term to long-term memory, build a custom memory store with embeddings, enabling chat summaries and user profiles.

Deploy memory for Azure AI agents by creating a memory store with a large language model and embedding model, then attach a memory search tool via code.

Set up a virtual environment, log in with Azure CLI, deploy AI Foundry resources with Bicep, and call a GPT 4.1 mini model via the OpenAI SDK.

Master Bicep deployment for Azure AI apps and agents by exploring tips, tricks, and practical advice in this special unit of the AI-103 course.

Explore how to tune large and small language model parameters, including max past messages, max tokens, temperature, and top P, via a config.yaml; assess impact on cost, latency, and routing.

Explore deploying embeddings with TextEmbedding3Small in Azure AI Search and Foundry, enabling hybrid and vector search with semantic ranking and a semantic configuration for re-ranking.

Deploy Azure AI Search resources with Foundry, content safety, and llms (large, small, embedding) in sequence, then enable grounding results for agentic rag to handle refund queries.

Explore tool calling with REST API integration, where AI agents automatically select, run, and ground tools, with content safety, intent classification, and model routing.

Explore the multi-agent handoff pattern that transfers conversation control between specialized agents using a handoff tool, conversation router, and separate agent classes for support and billing.

Learn the magentic manager pattern in this unit: a hierarchical multi-agent approach where a manager routes requests to specialized sub-agents (refund, products, accounts) with no state transfer.

Explore sequential and concurrent agent execution with LLM calls, including three classes for LLM calls, sequential, and concurrent, using asyncio for parallel tasks.

Deploy and use Azure Cosmos DB to store long-term user preferences, enabling cross-session memory for agents with memory, using a hybrid approach with Redis for short-term history.

Learn secure secret management with Azure Key Vault, deploying via Bicep, enabling purge protection and RBAC, using managed identities and a secret manager with caching to avoid hard-coded credentials.

Learn to build productionized AI systems on azure by integrating containers, vectors, and events. Use these pillars to enable scalable inference, semantic search, and decoupled workflows.

Learn how Azure Container Apps run container images from Azure Container Registry as serverless apps with autoscaling, revisions, environment variables, secrets, health probes, and KEDA-based scaling for AI services.

Build a basic FastAPI hello world API with get and post endpoints validated by Pydantic, including health checks, and deploy it to Azure Container Apps via a Docker image.

Build a configurable FastAPI app with the application factory pattern using createApp, include pydantic validation and endpoints, then deploy and test on Azure container registry and container apps.

Build an Azure Container Apps api that calls OpenAI chat completions via FastAPI, using the application factory pattern, an async client, environment-driven settings, and readiness probes.