RHEL AI: Best Practices and Optimization Techniques to Achieve Accurate Custom LLM

This conference talk by Deepak Das at DevConf.IN 2025 explores best practices and optimization techniques for achieving accurate custom Large Language Models using RHEL AI, a Red Hat product based on the upstream InstructLab project. Learn how to customize LLMs with private data while maintaining optimal performance through key optimization stages: Data Seed, Synthetic Data Generation, Training, Evaluation and Re-Training, and Prompt Engineering. Discover practical techniques applicable not only to RHEL AI but also to upstream InstructLab and third-party LLM models in general. The 49-minute presentation provides valuable insights for anyone looking to enhance LLM accuracy and performance in their AI implementations.