Optimizing TensorFlow model serving with Kubernetes and Amazon Elastic Inference Amazon Web Services