SnapperML: A python-based framework to improve machine learning operations Molner, Antonio Carrillo Pérez, Francisco Guillén Perales, Alberto MLOps Machine learning Deep learning Data Science has emerged as a vital discipline applicable across numerous industry sectors. However, achieving reproducibility in this field remains a challenging and unresolved problem. Additionally, transitioning trained models from development to production environments often proves to be a non-trivial task. In this study, we propose SnapperML, a comprehensive framework designed to address these issues by enabling practitioners to establish structured workflows that facilitate result reproducibility. Leveraging DevOps techniques, SnapperML ensures seamless model deployment from the lab to production, mitigating the risks associated with compatibility issues and model selection errors. The framework enables meticulous tracking of every aspect of model training, including hyperparameter selection, tuning, and distributed training on a server. By offering a suite of tools for model tracking and optimization, SnapperML presents a promising solution to the reproducibility challenge in the field of data science. 2024-05-14T10:46:05Z 2024-05-14T10:46:05Z 2024-03-04 journal article Antonio Molner et al. SoftwareX 26 (2024) 101648 [https://doi.org/10.1016/j.softx.2024.101648] https://hdl.handle.net/10481/91756 10.1016/j.softx.2024.101648 eng http://creativecommons.org/licenses/by-nc-nd/4.0/ open access Attribution-NonCommercial-NoDerivatives 4.0 Internacional Elsevier