From production LLM logs to better models

Production AI systems generate valuable data: prompts, responses, feedback, and interaction patterns. But inference logs usually sit in storage, disconnected from training, making model improvement slow and detached from real product behavior.Production AI systems generate valuable data: prompts, responses, feedback, and interaction patterns. But inference logs usually sit in storage, disconnected from training, making model improvement slow and detached from real product behavior.

In this session, we’ll show how teams turn production LLM logs into structured training datasets and use them to improve models through post-training.

You’ll see how to build a continuous improvement loop:
production inference → dataset → post-training → improved model → deployment

You’ll learn how toYou’ll learn how to

• Capture and explore production LLM inference logs
• Identify useful training examples from real usage
• Transform logs into structured training datasets
• Run post-training workflows on Nebius Token Factory
• Deploy improved models back into production
• Build a continuous improvement loop for LLM systems

Who should attendWho should attend

• ML engineers improving model performance with real usage data
• AI developers building copilots, assistants, and AI applications
• Platform teams designing data and training pipelines
• Founders and product leaders building AI-native products