Machine learning projects need more than a good model. A repeatable workflow for data, experimentation, evaluation, deployment and monitoring helps engineering teams manage the distinctive challenges of integrating ML into production software and systems.