24Offline the model scores 0.86 AUC. Served live, it's effectively random. Walk me through training/serving skew.▼hard★ EssentialGoogleDatabricksMicrosoft1 replies◆ premiumThe most expensive bug class in production ML: two implementations of 'the same' feature that quietly disagree. The diagnostic that finds it in an afternoon, and the architecture that makes it impossible.Open full answer →
16What is training-serving skew, and how do you keep online and offline features consistent?▼mediumGoogleUberMicrosoft2 replies○ sign inThe bug class with no stack trace: the model is fine, the pipelines are green, and production quietly underperforms offline by five points. Why skew survives even feature-store adoption, and the log-and-wait pattern Google-style answers center on.Open full answer →
