Talk: Evolution and Unification of Pinterest ML Platform (2021)

Talk at apply() Conference, April 21, 2021. (10 minutes)

As Pinterest grew over time, machine learning use cases proliferated organically across multiple teams, leading to a proliferation of technical approaches with bespoke infrastructure. The ML Platform team has been driving Pinterest Engineering to a unified platform to tame the complexity of these diverse use cases. This talk will give a brief history of the evolution of Pinterest ML and our layer-by-layer approach to standardization, including a unified feature representation, shared feature store, and standardized inference services. Finally, we share lessons learned in aligning multiple engineering orgs on a shared ML vision in the face of typical resource constraints and competing priorities.