Azure AI Foundry Model Catalog: Curriculum Overview
Describe features and capabilities of Azure AI Foundry model catalog
Azure AI Foundry Model Catalog: Curriculum Overview
This curriculum provides a structured path to mastering the Azure AI Foundry Model Catalog, a central hub for discovering, evaluating, and deploying foundation models within the Microsoft Azure ecosystem.
Prerequisites
Before engaging with the Azure AI Foundry Model Catalog, students should possess the following foundational knowledge:
- Azure Fundamentals: Basic understanding of cloud computing services, resource groups, and the Azure portal.
- AI Concepts: Familiarity with Artificial Intelligence terminology, specifically generative AI, foundation models, and machine learning.
- AI-900 Unit 1 & 2: Knowledge of responsible AI principles and basic machine learning techniques (regression, classification).
- Technical Environment: Access to an Azure subscription with permissions to create Azure AI Foundry projects.
Module Breakdown
| Module | Title | Focus Area | Difficulty |
|---|---|---|---|
| 1 | Foundations of AI Foundry | Overview of the platform (formerly Azure AI Studio) and project setup. | Beginner |
| 2 | Navigating the Catalog | Discovering models from Microsoft, Meta, Mistral, and Hugging Face. | Beginner |
| 3 | Model Evaluation | Benchmarking model performance and choosing the right size/type. | Intermediate |
| 4 | Deployment & Safety | Provisioning endpoints and configuring content filters/safety layers. | Intermediate |
Learning Objectives per Module
Module 1: Foundations of AI Foundry
- Define the role of Azure AI Foundry in the development lifecycle.
- Explain the transition from Azure AI Studio to Azure AI Foundry.
- Set up a project environment for model experimentation.
Module 2: Navigating the Catalog
- Identify the primary sources of models (OpenAI, Open Source, Partner models).
- Differentiate between model types (Text generation, Image generation, Embeddings).
- Understand the role of Hugging Face and other partners in the catalog ecosystem.
Module 3: Model Evaluation
- Utilize benchmarking tools to compare model performance on specific tasks.
- Analyze the trade-offs between model size (e.g., Llama 7B vs. 70B) and cost/latency.
- Review model cards for technical specifications and license constraints.
Module 4: Deployment & Safety
- Deploy models as "Serverless APIs" or to managed compute clusters.
- Configure Azure AI Content Safety layers to mitigate risks (harmful content, jailbreaks).
- Integrate deployed model endpoints into external applications.
Visual Overview
The Model Ecosystem
The Deployment Workflow
Success Metrics
To demonstrate mastery of the Azure AI Foundry Model Catalog, learners must be able to:
- Identify 5+ partners available in the catalog (e.g., Meta, Mistral, Anthropic).
- Explain the difference between Azure OpenAI models and Open Source models in the catalog.
- Describe the benefit of using a curated library (Security, Scalability, Governance).
- Select a model based on a specific scenario (e.g., choosing a small model for low-latency chat).
- Identify the safety layer used to filter inappropriate AI output in real-time.
Real-World Application
Understanding the Model Catalog is essential for professionals in the following roles:
[!TIP] Data Scientists use the catalog to quickly prototype with different architectures without managing complex local infrastructure.
[!IMPORTANT] Cloud Architects rely on the catalog's integration with Azure's secure, scalable infrastructure to ensure enterprise-grade compliance and data privacy.
Common Scenarios:
- Customer Support Chatbots: Deploying a specialized language model to handle natural language queries.
- Content Summarization: Using high-performance decoder blocks to turn long reports into actionable insights.
- Safety Filtering: Implementing real-time content moderation for user-generated prompts in a public-facing AI app.