Model Optimization in Microsoft Foundry: AI Agents Tool Calling Accuracy

Microsoft Developer presents a step-by-step video on improving AI agent tool-calling accuracy using fine-tuning in Microsoft (Azure) AI Foundry, including synthetic data generation and model distillation to produce smaller, more cost-effective models.

Overview

This video walks through improving tool calling accuracy for AI agents by using fine-tuning in Microsoft (Azure) AI Foundry (referred to as “Microsoft Foundry” in the description). It also introduces model distillation—taking the capabilities of a larger model and transferring them into a smaller model for lower cost and better deployability.

What you’ll learn

• How to improve reliable tool calling in AI agents
• How model distillation can shrink large-model capabilities into more nimble, cost-effective models
• A step-by-step flow covering:
  • Synthetic data generation
  • Advanced fine-tuning methodologies
  • Practical approaches to make agents more accurate

Video chapters

• 00:02 Welcome and scenario
• 03:05 Generating synthetic data
• 04:36 Model distillation and fine-tuning

Resources

• Microsoft Foundry: https://aka.ms/foundry-ft
• Foundry fine-tuning demos on GitHub: https://aka.ms/ft-demos

People

• Bethany Jepchumba
  • X/Twitter: https://twitter.com/bethanyjep
  • LinkedIn: https://www.linkedin.com/in/bethany-jep/
  • GitHub: https://github.com/bethanyjep