Abstract

Many of the features that a modern platform for machine learning requires, are features that regular statistical production can benefit from. Statistics Sweden, in collaboration with AI Sweden and Örebro University, is developing a platform called "BALSAM" to support machine learning initiatives as well as new ways of processing and analysing data.

BALSAM addresses both the technical and the non-technical challenges such as project templates, guidelines, data catalogues, capacity for reproducibility, versioning of data and code, and security mechanisms. The platform also enables collaboration between organizations and enables the pragmatic solutions necessary to ensure the success and sustainability of ML projects.
By using the same features for both regular statistical production as well as machine learning initiatives, BALSAM helps statistical organisations bridge the gap between statistical production and ML-ops.

This presentation will provide an overview of the BALSAM initiative and highlight how the platform can support statistical organizations in their transformation towards ML-ops while also enabling increased collaboration within the statistical community and beyond.