Deploying Kubeflow 1.3 RC with Argo CD

Kubeflow is a popular open-source Machine Learning platform that runs on Kubernetes. Kubeflow streamlines many valuable ML workflows i.e. it allows users to easily deploy development environments, scalable ML workflows with Kubeflow Pipelines, automated hyper-parameter tuning and neural architecture search with Katib, easy collaboration within teams and much more. With a such a large number of features also comes complexity. A full Kubeflow deployment contains many services and dependencies, making it difficult for users to customize, manage and install Kubeflow using the legacy kfctl CLI tool and a KfDef YAML file. For this reason, the upcoming Kubeflow 1.3 release has stopped using kfctl and instead is using standard Kustomize, making it easier to deploy Kubeflow with GitOps tools such as Argo CD.