Every ML person has chased a proxy that diverged from the thing that mattered. The screen is whether you noticed it yourself, why the proxy fooled you, and how you realigned without just swapping one number for another.
Updated Aug 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.
Every ML person has chased a proxy that diverged from the thing that mattered. The screen is whether you noticed it yourself, why the proxy fooled you, and how you realigned without just swapping one number for another.
The disqualifier is a story where someone else caught the problem and you executed their fix, because the question is testing whether you own the objective or just the loss function. The strongest version names the moment offline and online diverged, explains why the proxy was a reasonable choice that quietly stopped tracking the real goal, and shows you replacing it with something closer to the real objective rather than a different convenient proxy. Expect 'how do you pick a metric you won't have to walk back?', which probes whether you reason about proxy-target gaps before shipping.
No comments yet — be the first to share your approach.
