Background At Numerator, the Deep Learning Team hosts numerous models as a collections microservices on Kubernetes. We have a variety of building blocks in our architecture that allow us to build more complicated pipelines and processing steps. At the core, we have a fairly large number of pods that are