Machine Learning System Design Interview, issue 5, Nov 26, 2025
The Multi-Objective Loss Trap
ML Engineer interview at DoorDash, and the interviewer asks:
โProduct wants to maximize ๐๐ด๐ฆ๐ณ ๐๐ญ๐ช๐ค๐ฌ๐ด. Sales wants to maximize ๐๐ช๐จ๐ฉ-๐๐ฐ๐ฎ๐ฎ๐ช๐ด๐ด๐ช๐ฐ๐ฏ ๐๐ณ๐ฅ๐ฆ๐ณ๐ด. How do you design the ๐๐ฐ๐ด๐ด ๐๐ถ๐ฏ๐ค๐ต๐ช๐ฐ๐ฏ to balance these conflicting goals?โ
A mathematically elegant loss function can quietly destroy your real-time ranking system. Hereโs the architectural fix that actually works in production.
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.