Reading data and AI engines from the source code down.
Catalyst, scheduling, shuffle, memory, and codegen, traced through the Spark source.
Optimization techniques compared across engines: Spark, Velox, DataFusion, DuckDB, and more.