Mission to Maturity: Deploying MLOps to Scale Machine Learning at M&S by Sam Comber, Duong Le & Nick Masca

Constructing production-level machine learning (ML) systems invites an abundance of issues and obstacles absent from proof of concepts and small scale research projects. Data scientists typically implement and train ML models, yet the real challenge remains building integrated ML systems that operate continuously in production.

Practicing MLops advocates reproducibility, automated testing, and monitoring in the delivery of ML systems alongside checks such as data verification and model quality, allowing data scientists to release continuously and safely to production.

In this session, we explore how MLOps is deployed to scale ML at M&S across two diverse use cases that span from one-to-one personalisation which serves to enrich customer experiences, to enterprise functions that help M&S run better as a business. We demonstrate how projects at M&S follow the rubric of ML maturity, and exhibit how key testing and monitoring tools are deployed to ensure the production-readiness of our ML systems.

Supported by