MLOps

Abstract

Listen

Translate article

Companies are moving from a cottage industry to a factory approach to analytics, especially in regard to machine learning (ML) models. This change is motivating companies to adopt ML operations (MLOps) as a methodology for the timely development, deployment, and maintenance of ML models in order to positively impact business outcomes. The adoption of MLOps requires changes in processes, technology, and people, and these changes are informed by previous work on decision support systems (DSS), development operations (DevOps), and data operations (DataOps). The processes, technologies, and people needed for MLOps are discussed and illustrated using a customer purchase recommendation example. Current and future directions for MLOps practice driven by artificial intelligence (AI) are explored. Suggestions for further academic research are provided.