The document describes eBay's use of Spark and MLlib to build a scalable machine learning pipeline for metadata discovery from product listings. eBay developed supervised machine learning models using Spark MLlib to classify product listings and discover new metadata brands with high accuracy. The models were trained on historical metadata and product data, and can process large amounts of data quickly to discover new metadata on a daily basis, speeding up metadata discovery from months to days. Spark MLlib allows for efficient model development locally and simple deployment to production at scale.