​This session explores Apache Iceberg’s metadata landscape using agentic AI alongside Spark and Trino, demonstrating how different specification versions, including V3, influence performance and analytical capabilities. Viewers will see how AI can generate meaningful insights from Iceberg metadata and gain a practical view of what succeeds and what fails when handling large-scale and standard ingestion workloads. The session provides a grounded perspective on best practices, pitfalls, and lessons learned from real implementations.