Learn how to GPU-accelerate data pipelines, and how to apply that knowledge to build scalable recommender systems. In this session, we will discuss how to solve common bottlenecks found when building large-scale deep learning-based recommender systems, like long processing ETL times, and GPU underutilisation when loading data at training time. We will also learn how to efficiently train deep recommender systems, by making use of NVTabular and HugeCRT, two GPU-accelerated open-source libraries part of NVIDIA Merlin framework.