A Maturity Level Framework for Practicing Machine Learning Operations in CI/CD For Software Deployment

Abstract

Significantly shorter software development and deployment cycles have been made possible by the adoption of continuous software engineering techniques in business operations, such as DevOps (Development and Operations). Data scientists and operations teams have recently become more and more interested in a practice known as MLO (Machine Learning Operations). However, MLO adoption in practice is still in its early stages, and there aren't many established best practices for integrating it into current software development methods. In order to give a frame that outlines the way necessary in espousing MLO and the stages through which business capes process as they come riper and sophisticated, we achieve a methodical literature study as well as a slate review of literature in this composition. We test this approach in three example businesses and demonstrate how they were able to embrace and incorporate MLO into their massive software development businesses. This study offers three contributions. To give an overview of the state of the art in MLO, we first examine recent publications. Based on this analysis, we create an MLO framework that outlines the steps taken in the ongoing creation of machine learning models. Second, we define the several stages that businesses go through as they develop their MLO practices in a maturity model. Third, we map the firms to the maturity model phases and test our methodology using three embedded systems case companies. The main objective is to create an MLO framework.

Keywords

• Software DevOps
• Machine Learning
• Maturity Model
• Validation Study
• CI/CD
• Framework