The stack of deep learning is getting deeper and it's becoming hard for a single person to be proficient in every layer of the stack. The need for large amounts of compute is crucial as it enables researchers to make interesting discoveries, but it also presents the challenge of managing a huge compute cluster to run experiments.