SET OF RECOMMENDATIONS FOR IMPLEMENTING DEVOPS PRACTICES IN THE CONTEXT OF MACHINE LEARNING
DevOps practices are increasingly present in software engineering, aiming at the automation and simplification of processes that involve the application life cycle, from its development to deployment in a production environment. For machine learning, such practices are known as MLOps (Machine Learning Operations) and aim to integrate the stages of development and operations of ML systems, including automation and monitoring for the delivery and continuous updating of the system to the end user. Studies have shown the applicability of these practices in specific contexts in the literature. However, few provide guidelines to guide their implementation thoroughly, considering the entire set of factors involved in adopting MLOps – from cultural and organizational changes to the technical elements necessary for effective implementation. In this scenario, the present work proposes a set of recommendations to assist in the understanding and adoption of MLOps by researchers, managers, engineers, data scientists, and others involved in the development of machine learning systems. The recommendations are developed from a bibliographic review and a Design Science Research method to generate artifacts that will represent the detail and flow of the proposed recommendations. We also propose the application of a questionnaire to validate the applicability of these recommendations by professionals who work with ML systems development. We aim that the results found can serve as a reference source for researchers in the area and organizations.