Prestissimo Extension for AI Training Data Normalization at Meta

Learn about Meta's recent extension to Prestissimo that enables AI training data normalization through this 13-minute lightning talk. Discover how Meta built an AI Data Storage system designed to deduplicate user sequence data and facilitate rapid retrieval of aggregated data across multiple dimensions. Explore the technical modifications made to Prestissimo that allow user sequence data exploration via Presto SQL, including the implementation of a new Index Join Operator and AI Data Storage Connector. Understand how this extension achieves optimized index join query plan generation and comprehensive end-to-end query execution optimization for large-scale AI data processing workflows.