Although Nutrino had been successfully operationalizing ML models, the company decided to rearchitect their machine learning pipelines to comply with the FDA requirements related to model versioning and to automate manual deployment processes. To address their needs, they evaluated Kubeflow, MLFlow, Sagemaker, and AWS EFS. Ultimately, Gal Shen explains, Nutrino selected MLFlow as their model repository and artifact store and EFS to accelerate inference.