GO TO A SECTION

↑ ↓ to navigate · Enter to openExplore all sections →
FDEInterviews logo
← MLOps & ML Engineering / 16
mediumGoogleUberMicrosoft

What is training-serving skew, and how do you keep online and offline features consistent?

The 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.

Updated Aug 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.

The 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.

more free answers with an account · no card

Keep practising

Next in this trackDesign a feature store for a real-time fraud detection model.Free with sign-inNext in this trackWhich statistical tests would you use to detect drift, and what thresholds should trigger action?Free with sign-inNext in this trackHow do you monitor a model in production when ground-truth labels arrive late, or never?Free with sign-in
DISCUSSION · 0

No comments yet — be the first to share your approach.