How to Monitor Machine Learning Stacks - Detecting and Addressing Special Issues

Discover effective strategies for monitoring machine learning stacks in this insightful conference talk by Lina Weichbrodt, Machine Learning Lead Engineer at DKB Bank. Learn why traditional monitoring methods often fail to detect critical issues in ML services and explore innovative solutions to address these challenges. Delve into practical examples that illustrate the unique problems faced by machine learning systems and gain knowledge about additional metrics that can be implemented to enhance detection capabilities. Examine a compelling case study from Zalando, one of Europe's leading fashion retailers, showcasing how these new metrics are applied to their recommendation stacks. Gain valuable insights into improving the reliability and performance of machine learning services beyond the conventional "four golden signals" of latency, errors, traffic, and saturation.