Build custom AI app experiences grounded on your own data in Azure AI Foundry — deploy models like GPT-4o, add Retrieval Augmented Generation with vector search, orchestrate multi-step agents, and run evaluations and monitoring as part of GenAIOps.
Azure AI Foundry gives developers 1,800+ models, prebuilt Azure AI services, and end-to-end tooling to build, evaluate, and monitor generative AI apps directly from your code.
👥 Who it's for: Developers and AI engineers building generative AI apps, Azure developers choosing and deploying models, and teams standing up GenAIOps for evaluation and production monitoring.
⏱️ Chapters:
00:00 Build custom AI apps in Azure AI Foundry
00:27 Model catalog & benchmarks to pick a model
01:37 Ground apps on your own custom data
03:03 Add Retrieval Augmented Generation (RAG)
03:48 Create a vector search index with hybrid search
04:17 Set up your code with the OpenAI SDK
06:11 Orchestrate multi-step AI agents in code
07:16 Debug model calls with traces
07:45 Evaluate quality with Azure AI evaluators
08:19 Monitor production apps in Application Insights
08:58 What to build first in Azure AI Foundry
Azure AI Foundry (now generally available) brings the studio and Azure AI together so you can go from picking a model to shipping a grounded, production-ready app in code. Start in the model catalog to browse 1,800+ models from OpenAI, Microsoft, Meta, Mistral and more, then use model benchmarks — accuracy, coherence, groundedness, fluency, and relevance — to choose the right one. Prebuilt Azure AI services add cognitive skills like speech, language and translation, vision and OCR, and content safety to detect harmful inputs and outputs.
Dan Taylor, Principal Product Architect for the Azure AI Foundry SDK, shows how to create a project, deploy GPT-4o, and set a system message in the playground. To ground responses on your data, he adds Retrieval Augmented Generation (RAG), uploading product files to automatically build an Azure AI Search vector index with hybrid keyword + vector search. From there, he moves to VS Code using the OpenAI SDK and Azure Identity — calling Azure OpenAI with your Azure credential and no API keys.
The video then builds a multi-step orchestration where different models and prompty prompt templates run research, product, assignment, and editor tasks to write an article. Using azd up and app templates, GitHub pipeline config, traces for debugging, and built-in Azure AI evaluators (scored 1–5 on relevance, fluency, coherence, and groundedness), you can test, optimize, and monitor everything in Application Insights — token usage, model duration, evaluation scores, and transaction-level trace search.
► Link References
- Find Azure AI Foundry at https://ai.azure.com
- Check out our code samples at https://aka.ms/AIAppTemplates
► Unfamiliar with Microsoft Mechanics? Microsoft's Official Video Series for IT
- Subscribe https://www.youtube.com/c/MicrosoftMechanicsSeries
- Microsoft Tech Community: https://techcommunity.microsoft.com/t5/microsoft-mechanics-blog/bg-p/MicrosoftMechanicsBlog
- Podcast: https://microsoftmechanics.libsyn.com/podcast
► Join us on social:
- https://twitter.com/MSFTMechanics
- https://www.linkedin.com/company/microsoft-mechanics/
- https://www.instagram.com/msftmechanics/
- https://www.tiktok.com/@msftmechanics
#AzureAIFoundry #AzureAI #GenerativeAI #Python #GPT4o