Comparative Study of Open-Source CI/CD Tools for Machine Learning Deployment

Abstract

The adoption of Continuous Integration and Continuous Deployment (CI/CD) tools has transformed the landscape of machine learning (ML) workflows, enabling automation, scalability, and efficiency. This study evaluates the comparative performance of three prominent open-source CI/CD tools\u2014Jenkins, GitHub Actions, and Bitbucket Pipelines\u2014in addressing the unique demands of ML tasks, including hyperparameter tuning, model training, and deployment. Through a systematic analysis, the research explores key parameters such as scalability, usability, and security integration, providing actionable insights into their suitability for diverse organizational contexts. Jenkins, with its extensive customization options, demonstrates flexibility but is hindered by a steep learning curve. GitHub Actions excels in usability and accessibility for smaller teams but requires enhancements to handle large-scale workflows. Bitbucket Pipelines, with Kubernetes integration, emerges as a robust option for resource-intensive tasks, though its documentation and advanced features need refinement. The study highlights critical gaps in existing tools, such as limited scalability for distributed workloads and insufficient integration of advanced security mechanisms like TLS automation. Recommendations for tool selection and future enhancements are provided, emphasizing adaptive pipelines, federated learning workflows, and energy-efficient orchestration. This work contributes to the optimization of CI/CD tools for ML operations, offering a structured framework and practical guidance for practitioners and researchers aiming to deploy secure, scalable, and efficient ML pipelines