This is a joint talk given by CashApp and Tecton. CashApp’s mobile payment product has stringent technical requirements: scale, reliability, speed. ML-based recommendations are at the core of this service and pose a significant engineering challenge. This talk describes CashApp’s journey through various generations of its core ML capabilities, covering the technical and organizational challenges associated with building large-scale production recommendation systems. The talk finishes with a look at the latest generation of CashApp’s ML platform and highlights how Tecton’s real-time Feature Platform helps CashApp deliver world-class recommendations.