Booking.com: Building a Scalable Machine Learning Platform
"The H2O integration with Spark gives you the best of both worlds: Data munching in Spark and model training in H2O."
Senior Data Scientist
Scaling ML Platform
Overview of the Challenge
Brammert talks about how to build a machine learning platform that scales to support Booking’s 200 data scientists and 1.5 nights reserved every day. He explains the ML pipeline from data collection, training, to production including model and feature monitoring and discoverability. He shows how Booking.com is able to scale their experiments to billions of rows of data using H2O and Sparkling Water.