In the past few months, several data practitioners have presented and discussed the concept of a “data mesh” wherein data is treated as a product that is managed in a decentralized way, including by data engineers embedded within product engineering teams – thereby, improving data quality and governance. However, few have discussed the technical implementation of data mesh. In this post, Netflix engineers discuss the technical architecture of the data mesh that enables them to provide real-time operational insights to various consumers of Netflix Studio data. Their data mesh hinges largely on change data capture (CDC) systems that are used to sync data between sources and sinks. Users can define these pipelines through a self-service user interface, including by applying reusable processors. The Netflix data mesh also supports schema evolution and efficient onboarding of new entities.