Utilization of Kubeflow for Deploying Machine Learning Models Across Several Cloud Providers

Kubeflow is a global machine learning platform that is open-source in nature. Its primary objective is to streamline the process of deploying, managing, and scaling machine learning (ML) and deep learning (DL) workflows. This platform is specifically built to operate on Kubernetes, which is a widely used container orchestration platform. Kubeflow offers a comprehensive platform for managing machine learning operations (MLOps), facilitating efficient collaboration across data scientists, ML engineers, and DevOps teams at every stage of the machine learning lifecycle. The use of Kubernetes' containerization capabilities facilitates the establishment of a portable and scalable environment specifically designed for innovative machine learning applications. The objective of this study is to investigate the procedure of installing Machine Learning models on Kubernetes by using Kubeflow, an open-source technology recognised as an end-to-end ML Stack orchestration toolkit. In this study, development of comprehensive Machine Learning models is done using Kubeflow in the form of pipelines. Then many aspects of the tool are analysed, including the simplicity of its setup, the deployment of models, their performance, restrictions, and features. This project aims to serve as a seminar or introduction report, providing assistance to vanilla cloud/Kubernetes users who possess no prior knowledge of Kubeflow. The objective is to guide them in using Kubeflow for the deployment of machine learning models. This research provides comprehensive information and analytics on the performance of Kubeflow, including many aspects such as the setup on multiple cloud platforms and the deployment of our trained model for serving over the internet.