The paper presents Aster, a system designed to simplify the deployment and management of machine learning models. The authors propose a scalable and flexible architecture for model serving, which enables efficient model deployment and updates.
