Inside AIStor Tables: How to Unify All Your Data for AI Workloads

Enterprises today face a familiar challenge: their most valuable data (structured and unstructured) remains siloed across systems. For AI agents and analytics to reach their full potential, all data must be discoverable, queryable, and governed together.

Join Brenna Buuck, Developer Evangelist at MinIO, for a live technical deep dive into AIStor Tables, a new feature that makes the Iceberg REST catalog native in AIStor.

In this hands-on session, Brenna will demonstrate a manufacturing quality control AI agent use case that showcases the real power of unified data access:

• Pointer-based table design: Create Iceberg tables where structured metadata (sensor readings, batch numbers, defect rates) lives alongside S3 pointers to unstructured assets (inspection images, maintenance logs, video footage). In production architectures, these pointers would extend to vector embeddings in your vector database of choice.
• Smart filtering patterns: Watch how AI agents query structured metadata first ("find all batches with defect rates >2%") then retrieve only the relevant images or videos for analysis, avoiding the costly pattern of scanning entire AIStor buckets.
• ACID guarantees at scale: Experience how Iceberg's transactional semantics ensure metadata and documents stay in sync, even as your AI agents make real-time decisions.
• Multi-engine interoperability: Query the same tables with PyIceberg for ML pipelines, Spark for batch updates, and Trino for real-time analytics (all without moving data).

Why This Matters: Traditional architectures force you to choose: either scan millions of files (slow, expensive) or maintain separate metadata systems (complex, fragile). AIStor Tables eliminates this tradeoff by treating objects and tables as first-class citizens in the same storage layer, enabling AI agents to intelligently traverse both structured and unstructured data through native pointer-based schemas. While this demo uses direct S3 document pointers to illustrate the core concept, the same architecture pattern extends naturally to production RAG systems where your Iceberg tables maintain pointers to vector embeddings, chunked documents, and their metadata across distributed vector databases.

Who should attend: Whether you’re a developer, data architect, or platform engineer, this session will show how MinIO AIStor Tables redefine what’s possible for on-premise, AI-driven data architectures.