MLOps Tools for Deployment: A Case Study on Text Classification

Abstract

This paper explores the integration of MLOps tools in the design and operationalization of machine learning pipelines for text classification. Focusing on a case study of web news classification, we examine the use of tools such as MLflow for experiment tracking, Docker for containerization, and Airflow for orchestration. The results reveal both promising advancements and significant limitations, underscoring the challenges of adopting DevOps practices in the rapidly evolving field of machine learning. Although the findings highlight the potential of MLOps to improve scalability and reproducibility, they also demonstrate that the domain is still in its early stages and requires further refinement.