A Research On MLOPS Automation Pipelines With Kubeflow Stack And VSCODE Remote Development

Abstract

The growing involvement of machine learning in businesses brings both advantages and challenges. One of the biggest challenges is what it take to actually bring a model to production. MLOps, inspired by the already famous DevOps concept widely adopted in Software Engineering, appeared as the solution to the problem. Since then, many projects have sprung up and compete each other to provide a proof of concept for MLOps. Most are young and immature, but some achieved huge success and skyrocketed in popularity. This thesis will point out some of the serious issues in one of the best and most popular MLOps platform at the time and then make some contributions to it in order to overcome those issues and then package them as my own distribution of the tool.