Microsoft Reactor

Modern AI systems require more than powerful models—they require security, evaluation, governance, and continuous improvement. This session combines lessons from production AI agent security with real-world LLM evaluation and fine-tuning workflows.

Topics may include prompt injection, tool abuse, memory poisoning, defense-in-depth architectures, custom evaluation frameworks, Azure OpenAI fine- tuning, and practical engineering lessons learned from deploying AI-powered systems.

Key Takeaways:

• Understand security challenges in AI agents
• Learn practical defense patterns for production AI
• Explore LLM evaluation methodologies
• Understand fine-tuning workflows using Azure OpenAI
• Apply production engineering best practices to AI systems