BigML pioneered ML platforms back in the day: it was 2011 and only Google's blackbox predictions were offering something similar to an API-based ML service. It was also the hype time of Big Data and Hadoop, when everybody first installed a big cluster with complex infrastructure and, afterwards, wondered what problem they were going to solve. This is the story of how we side-stepped the hype, kept our small start-up focused, and faced the many technical challenges of building a scalable, easy-to-use, API-first, DSL-based platform aimed at making ML accessible to everyone.