AI Fine-Tuning in Microsoft Foundry: Building Production-Ready Agents
Microsoft Events presents an advanced demo-driven session featuring Alicia Frame, Ramachandra Kota, and Omkar More. They explore fine-tuning in Microsoft Foundry for Azure OpenAI and open-source models, highlighting synthetic data strategies and AI-powered agent development.
AI Fine-Tuning in Microsoft Foundry: Building Production-Ready Agents
Session Overview
This Microsoft Ignite breakout session explores how fine-tuning in Microsoft Foundry enables the creation of agents that excel in practical scenarios. The session covers:
- Model selection and customization (Azure OpenAI, OSS)
- Synthetic data generation for training
- Tool calling, data extraction, and workflow execution
- Multi-agent strategies for high-quality results
- Real-world applications, including contract repository solutions
Key Topics and Demos
1. Practical Model Fine-Tuning
- Steps: choosing models, generating relevant training data, evaluating model performance.
- Announcement: Microsoft Foundry now supports automatic synthetic data generation for training, simplifying the data preparation process.
2. Synthetic Data Generation
- Utilizes Swagger definitions to guide automatic generation of quality data.
- Demonstrated multi-agent collaboration to improve synthetic data quality.
- Example: Creating diverse scenarios for agent training.
3. Advanced Agent Capabilities
- Agents are designed for accuracy and reliability in production.
- Demo: Foundry launching RFT using models like GPT-5 and O4 Mini for complex query handling.
4. Real-World Application: AI-Powered Contract Repository
- ‘Navigator’ is introduced as an AI-powered solution for storing and processing contracts.
- Data extraction: AI identifies key contract details and relationships, enabling the processing of millions of documents daily.
5. Scalable AI Workflows
- Focus on building agents that perform, not just respond.
- Covers workflow execution and best practices for scaling agents in enterprise environments.
Speaker Information
- Alicia Frame
- Ramachandra Kota
- Omkar More
Additional Resources
Summary
This session empowers technical practitioners to leverage Microsoft Foundry and Azure OpenAI for advanced model fine-tuning, synthetic data generation, and deploying robust, scalable AI agents in production workflows.