In recent years, statisticians trained in data science methods have been joining National Statistical Offices (NSOs), aiming to harness nontraditional data sources and machine learning methods to enhance the production of official statistics. Despite their expertise, these professionals encounter significant barriers, including limited computational resources, inflexible development environments that do not foster collaborative work, and limited tools to transition from innovative experiments to production-grade solutions. This chapter presents Onyxia, an open-source project developed to address these challenges by enabling organizations to build modern and flexible data science environments that enhance the autonomy of statisticians. With Onyxia and its showcase instance, the SSP Cloud, we demonstrate how cloud technologies can be made readily accessible, fostering innovation, collaboration, and reproducibility in the field of official statistics. Through a case study of the classification of the activity domain (NACE) of French companies, we illustrate how these tools have been instrumental in operationalizing machine learning models in accordance with MLOps principles, marking a significant step forward in the valorization of data science projects at INSEE.