Azure AI Foundry
Azure AI Foundry (formerly Azure AI Studio) is Microsoft’s integrated platform for designing, customizing, and managing AI applications and agents.
Overview
Go to Azure AI Foundry portal from Azure portal (like Azure OpenAI Service, or previous Azure AI Studio). You can compare models in the catalog with quality and cost.
AI services: Model Catalog, Azure AI language/speech/vision, content understanding/document intelligence
Data Services: Azure AI Search, Vector store
Development: Prompt Flow
MLOps/LLMOps
Azure AI model inference service
Azure AI agent services
Azure AI Content Safety
Evaluation, Tracing
Azure AI services are AI services provided (e.g. OpenAI models, Cognitive services). However, AI application need to have other things that build on them. How they are managed?
An Azure AI Foundry project is where you do most of your development work. You can work with your project in the Azure AI Foundry, or using the SDK in your preferred development environment. Once you have a project, you can connect to it from your code. You can explore models and capabilities before creating a project, but once you’re ready to build, customize, test, and operationalize, a project is where you’ll want to be.
At a glance, Azure AI services UI and Azure AI Foundry project UI are very similar (hope this can be improved and more intuitive).
Projects live inside a hub. A hub allows you to share configurations like data connections with all projects, and to centrally manage security settings and spend. If you’re part of a team, hubs are shared across other team members in your subscription. It provides the following features:
- Security configuration including a managed network that spans projects and model endpoints.
- Compute resources for interactive development, fine-tuning, open source, and serverless model deployments.
- Connections to other Azure services such as Azure OpenAI, Azure AI services, and Azure AI Search. Hub-scoped connections are shared with projects created from the hub.
- Project management. A hub can have multiple child projects.
- An associated Azure storage account for data upload and artifact storage.
Github marketplace for models, fine-tuning for models and evaluation, Azure AI agent service
RAG
multimodal rag with llamaparse, llamaindex and Azure AI search
Agents
Knowledge (data sources), action (logicapp, azure function, OpenAPI, function calling, code interpreter)
