The Models Going Stale

A hard Pipeline Design interview practice problem on DataDriven. Write and execute real pipeline design code with instant feedback.

Domain

Pipeline Design

Difficulty

hard

Seniority

L6

Problem

Our risk models keep going stale in production: the features behind them aren't refreshed fast enough to score loan applications live, and the models don't all tolerate the same lag. Design a feature pipeline and store that serves fresh features at low latency for live scoring while keeping years of historical feature values, so the team can retrain on what was true at the moment of each past decision. Compliance also needs to trace exactly which feature values produced any given score.

Summary

The model is only as good as what you feed it.

Practice This Problem

Solve this Pipeline Design problem with real code execution. DataDriven runs your solution and checks it automatically.