AI 900 정리 - Azure

Azure AI

• Can be provisioned as standalone service
  • Ex. Azure AI Vision
  • Azure AI Language
• Or provision a multi-service Azure AI service that encapsulate everything
• Some of them are available in only subset of azure region

Azure AI Foundry

Hub

• Provide centrally managed collection of shared resources and management configuration for AI solution development

Projects

• A hub can support one or more project

A unified platform for quickly building and deploying AI applications, offering the following key capabilities:

1. 📦 Model Catalog

   • Discover and deploy machine learning models from multiple sources:
     • Azure OpenAI
     • Hugging Face model library

2. 🧪 Prompt Testing Playgrounds

   • Interactive environments to experiment with prompts using generative AI models

3. 🔧 Access to Azure AI Services

   • Visual interfaces to configure and test services
   • Endpoints and keys available for integration with client applications

4. 💻 Visual Studio Code Containers

   • Hosted development environments for writing, testing, and deploying code
   • Preconfigured with necessary tools and dependencies

5. 🎯 Fine-tuning Capabilities

   • Customize generative AI models with your own training prompts and responses

6. 🔁 Prompt Flow

   • A prompt orchestration tool for defining the interaction logic between your app and AI models

7. 🧠 Evaluation and Monitoring Tools

   • Tools to assess, trace, and improve AI applications
   • Includes evaluations, tracing, and content safety/security features

8. 📂 Project Asset Management

   • Centralized management of:
     • Models and endpoints
     • Data and indexes
     • Deployed web apps