Creating Mollie’s ML Platform: from notebook to production with Vertex AI

Going to production has traditionally always been a big deal, especially for Machine Learning models. It generally involved many people with different roles, lots of hand-overs between these people, and sometimes even rewrites of the existing code. The practice of MLOps aims to address all of these by applying tried-and-tested DevOps principles to Machine Learning. At Mollie, we have fully embraced this methodology and have setup an ML Platform using Google’s Vertex AI. In this talk, we’ll discuss what this means in practice by showing you how we take a model from an idea and initial implementation to production. We’ll show how we iterated over various approaches, and how we’ve customized Google’s Vertex AI offering to suit the needs of Mollie.