Watch this video on YouTube
Which in-memory columnar data format is used by Pandas API on Spark to efficiently transfer data between JVM and Python processes?