Debug online worse than offline model performance
Company: Amazon
Role: Applied Scientist
Category: ML System Design
Difficulty: medium
Interview Round: Onsite
Overview: This question evaluates the ability to diagnose discrepancies between offline and online model performance by reasoning about data distributions, feature computation, serving infrastructure, evaluation methodology, and feedback loops, testing skills in debugging model deployment and operational analytics.
Read the full Amazon Applied Scientist interview experience this question came from
