Fine-tuning LLMs for Multi-Agent Orchestration - Cosine AI Case Study

Explore advanced techniques for optimizing large language models in multi-agent AI systems through this AWS re:Invent 2025 conference session. Discover how to implement fine-tuning, knowledge distillation, and preference alignment methods to create specialized models that are small, fast, and cost-effective for production applications. Learn from Cosine AI's real-world case study as they demonstrate how they leverage multi-agent reasoning to power their AI coding agent and scale customized agents effectively. Examine AWS-based multi-agent systems that enable continuous improvement of fine-tuned LLMs with minimal human intervention, focusing on robust coordination between specialized agents. Gain insights into reducing latency and costs while maintaining performance in multi-step production applications, and understand how these techniques enable effective orchestration of AI agents for real-world deployment scenarios.