retkowsky/Azure-AIGEN-demos

What it solves

This repository provides a practical, hands-on collection of Jupyter notebooks that demonstrate how to build AI-powered applications using Microsoft Foundry (Azure AI Foundry) and Azure OpenAI. It addresses the need for concrete, runnable examples that cover everything from basic model usage to advanced multi-agent orchestration, image and video generation, document analysis, and AI system evaluation. Instead of searching through scattered documentation, developers get a single place with working code for common and cutting-edge scenarios.

How it works

The project is organized as a set of Jupyter notebooks, each focusing on a specific capability. The notebooks are self-contained and include code samples that you can adapt and run in your own Azure environment. The demos cover a wide spectrum: from simple text generation with models like GPT-4o and GPT-5, to creating images with tools like gpt-image-1 and FLUX, to building agents with Azure AI Agent Service (including single-agent, multi-agent, Bing search integration, and MCP connections). It also includes notebooks for fine-tuning models (e.g., GPT-4o for image classification), for document intelligence with Mistral, for embedding-based search and RAG patterns, and for evaluating AI outputs using built-in and custom evaluators. Many notebooks include links to the Azure AI Foundry portal, so you can see the results in a managed environment.

Who it’s for

This is aimed at developers, data scientists, and AI practitioners who want to learn Microsoft's Azure AI ecosystem or need a rapid starting point for building enterprise-grade AI applications. It’s especially useful for those who are new to Microsoft Foundry or want to explore advanced topics like agentic workflows, multimodal generation, and AI observability. The demos serve as both educational material and a source of ready-to-adapt code for real projects.

Highlights

• Broad coverage: from basics of Azure OpenAI to advanced topics like Prompt Agents, MCP, and model routing.
• Multimodal demos: image generation with gpt-image-1, gpt-image-2, FLUX, SORA (video), and audio generation.
• Agentic scenarios: single-agent, multi-agent orchestration, function calling, and custom Bing integration.
• RAG and retrieval: simple file search, agentic retrieval in Azure AI Search, and embeddings.
• Evaluation and tracing: notebooks for using built-in evaluators, custom evaluators, and model tracing.
• Fine-tuning examples: GPT-4o fine-tuning for image classification and visual question answering.
• Document analysis: PDF parsing, resume analysis, and document AI with Mistral and OCR.
• Practical applications: podcast generation, PowerPoint creation, and text-to-emoji classification.