Design & Implementation of MLOps (Machine Learning Operations) Platform in Cloud Environment

Abstract

Machine Learning Operations (MLOps) concept has emerged basically to decode the shortcomings of traditional Machine Learning (ML) life cycle management in industrial applications. The incorporation of Continuous Integration (CI), Continuous Deployment (CD) and Continuous Training (CT) throughout different phases of MLOps facilitates improved ML model performance on recurrent basis. In this master’s thesis, at first, we delineate every single concept of MLOps, then determine the appropriate tools and services in a conscientious way to design organization centric MLOps solution architecture and finally implement it as a unified resolution in cloud environment. This end-to-end automated workflow implementation to orchestrate the whole ML lifecycle is a pivotal step to standardize the MLOps practice for the successful endeavour of ML products in business.