Developing and Deploying ML Across Teams with MLOps Automation Tool

We developed an MLOps automation tool in order to prevent ML engineers working on different projects from continuously repeating the same devops tasks manually. We standardized the management of cloud resources using infrastructure as code as well as model training and deployment workflows using best practices as code. Cloud agnostic infrastructure creation is performed using infrastructure as code templates rendered by the automation tool and Terraform, models are trained using MLflow with custom Kubernetes-backend plugins and a custom pipeline executor, models are deployed and monitored using rendered templates and Seldon-Core. Our MLOps tool allows machine learning engineers to perform DevOps tasks they were not trained to perform, standardize different projects, and save time and cost by reducing repeated work.