Azure AI Foundry Cheat Sheet 

Azure AI Foundry is a unified platform that enables enterprises to design, customize, and manage AI applications and agents at scale. It integrates tools, models, and workflows to streamline the development and deployment of AI solutions.

Key Components

• AI Foundry: A development environment for building, testing, and deploying AI models and applications.

• Model Catalog: A repository of prebuilt models from Microsoft, OpenAI, and other partners, facilitating model selection and deployment.

• Prompt Flow: A tool for designing and orchestrating language model workflows, enabling systematic experimentation and refinement.

• Agent Service: A platform for securely designing, deploying, and scaling AI agents, integrating with various data sources and action connectors.

• Management Center: A centralized dashboard for monitoring and managing AI resources, ensuring governance and compliance.

Features & Capabilities

• Model Customization: Supports fine-tuning of models from Azure OpenAI Service, Phi family, and over 1,600 models in the catalog.

• Data Ingestion: Integrates with various data sources, including Azure AI Search, SharePoint, and Microsoft Fabric, for comprehensive data access.

• Multi-Model Orchestration: Facilitates the combination of different AI models to solve complex tasks more effectively.

• Security & Compliance: Offers enterprise-grade security features and compliance certifications, including ISO/IEC 42001:2023.

• Observability: Provides tools for monitoring AI performance, tracking metrics, and ensuring responsible AI practices.

Use Cases

• Customer Support Automation: Develop AI agents to handle customer inquiries and support tasks.

• Document Generation: Automate the creation of contracts, invoices, and other documents by extracting and summarizing relevant information.

• Knowledge Mining: Extract insights from large datasets to inform decision-making processes.

• Generative Content Creation: Generate product descriptions, marketing materials, and other content using AI models.

Supported Models & Tools

• Proprietary Models: Azure OpenAI Service models, including GPT-4 and GPT-3.5.

• Open-Source Models: Models from Hugging Face, Meta, and other partners.

• Azure OpenAI: Access to foundational and reasoning models, with customization capabilities.

• LangChain & Semantic Kernel: Libraries for building and orchestrating language model applications.

• Hugging Face on Azure: Deploy tens of thousands of pretrained Hugging Face Transformers models in Azure Machine Learning using Hugging Face endpoints.

Deployment & Governance

• Deployment Options: Supports deployment on Azure Kubernetes Service (AKS), Azure Functions, and containerized environments.

• Scalability: Offers scalable infrastructure to handle varying workloads and ensure high availability.

• Governance Tools: Integrates with Microsoft Purview, Credo AI, and Saidot for tracking model performance, fairness, and regulatory requirements.

How to Get Started

1. Create an Azure Subscription: Sign up for an Azure account if you don’t have one.

2. Set Up AI Foundry: Navigate to the Azure portal and create an Azure AI Foundry resource.

3. Access AI Studio: Use AI Studio to begin building and deploying AI models and applications.

4. Explore Model Catalog: Browse the Model Catalog to select appropriate models for your use case.

5. Integrate Tools: Utilize Prompt Flow, Agent Service, and other tools to design and orchestrate AI workflows.

Licensing / Pricing Model

Azure AI Foundry offers a pay-as-you-go pricing model, with costs based on resource usage, model deployment, and data processing. For detailed pricing information, refer to the Azure AI Foundry pricing.

Limitations / Considerations

• Region Availability: Some features may not be available in all Azure regions.

• Model Limitations: Certain models may have specific usage restrictions or licensing terms.

• Data Privacy: Ensure compliance with data privacy regulations when handling sensitive information.

• Resource Management: Monitor and manage resources to optimize costs and performance.

References:

• https://azure.microsoft.com/en-us/blog/new-capabilities-in-azure-ai-foundry-to-build-advanced-agentic-applications/
• https://azure.microsoft.com/en-us/products/ai-model-catalog
• https://azure.microsoft.com/en-us/products/ai-agent-service
• https://azure.microsoft.com/en-us/blog/azure-ai-foundry-your-ai-app-and-agent-factory/