Presented by Microsoft Events and Katherin Madche at Ignite 2025, this session demonstrates how to build and evaluate robust AI agents using Foundry Agent Service, Azure Functions, and W&B Weave.

Building and Evaluating AI Agents with Foundry Agent Service

Speaker: Katherin Madche
Event: Microsoft Ignite 2025
Session Code: ODSP1411
Language: English (US)

Overview

Building intelligent AI agents requires robust tools and seamless integration. This session demonstrates how the Microsoft Foundry Agent Service can accelerate agent creation and deployment with the security, scalability, and flexibility of Azure services.

What You'll Learn

• How to prepare a Function App in Azure with API keys.
• Registering tools and connecting live data endpoints for agent use.
• Integrating Azure Functions as real APIs with agents.
• Using W&B Weave for tracking agent reasoning, input/output, and logging.
• Step-by-step Python-based interaction with your agent.
• Testing and evaluating agent output (e.g., live data queries, such as finding nearby restaurants).

Session Chapters

• 0:00 – Preparing Function App with API Keys
• 3:16 – Functions Live for Agent Use
• 4:38 – Tools Introduction and Dashboard Overview
• 6:09 – Testing Agent Using ‘Try in Playground’ Interface
• 7:40 – Viewing Trace Logs in Weave
• 8:09 – Displaying Agent Output: Restaurant Results Example

Key Integration Points

• Foundry Agent Service: Manage, scale, and secure cloud-based AI agents.
• Azure Functions: Deploy APIs for agent consumption.
• W&B Weave: Track agent workflows, inspect logging, and monitor real-time analytics.
• Python SDKs: Scripting and interaction with the agent.

Resources

• Building and evaluating AI agents with Azure AI Agent Service and W&B Weave
• Microsoft Ignite On-Demand Sessions

About the Presenter

Katherin Madche shares hands-on technical guidance in a practical developer-oriented format, enabling attendees to design, build, and troubleshoot Microsoft-based intelligent agents for a wide range of scenarios.