Pinterest Machine Learning Engineer Interview Questions

Pinterest·Machine Learning Engineer·Onsite - System Design / Architecture

Jul 2026

Pinterest ML Engineer interview focused entirely on designing an ad ranking system for a personalized feed. The session was deep and technical, with the interviewer pushing hard on training data construction, bias handling, and embedding strategies. Not a casual screen.

Design an ML system to increase click-through rate for ads shown in a personalized social media feed. Walk through how you'd frame the problem, build and train the model, serve it, and operate it end to end.
What metrics would you track for this system, both online and offline? How do you prevent CTR optimization from degrading user experience or long-term engagement?
How do you construct training data for the CTR model? What are your positive and negative examples, and what logs do you need?
How would you build user and ad embeddings? Describe both a two-tower retrieval approach and a graph-based approach, including what features go into each and how you'd keep them fresh.
How do you handle position bias and selection bias in your training data, and what's your approach to calibration given that the model output feeds directly into auction scoring?
Your model shows strong offline AUC improvements but the A/B test comes back flat or negative. What are the most likely causes and how do you diagnose them?
How would you jointly optimize for CTR and post-click conversions without one signal cannibalizing the other? How does the auction score change in that setup?
A brand-new advertiser campaign launches with zero history. Trace exactly how it gets retrieved, ranked, and given a fair opportunity to accumulate signal without burning through budget.
How do delayed conversions break a naive training pipeline, and what changes would you make to handle them properly?

“This is the main question and it's basically a full system design session disguised as one prompt.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jul 2026

Pinterest ML Engineer interview that was basically a full coding challenge to implement Bagging with Decision Trees from scratch, no numpy allowed. Pure Python only, which sounds fine until you're rebuilding Gini impurity with nested lists at 11pm.

Implement Bagging (Bootstrap Aggregating) with Decision Trees entirely from scratch in pure Python, no numpy. You need to implement bootstrapping with replacement, fit multiple trees on bootstrap samples, and aggregate predictions via majority vote for classification or mean for regression.
How do you handle the prediction aggregation step in your Bagging implementation, specifically for both classification and regression cases?
What stopping criteria did you implement for your Decision Tree, and why did you choose those specific ones?

“This one took me a while to even scope properly.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jul 2026

Pinterest ML Engineer interview with a coding round that leaned more algorithmic than I expected. One problem stood out and I'm still thinking about whether my approach was actually clean or just lucky.

You're given an integer array called target. Starting from an all-zeros array of the same length, each operation lets you pick any subarray and increment every element in it by 1. What's the minimum number of operations needed to reach the target array?

“The key is realizing each operation maps to a contiguous range of increments, so you're really counting how many times the value goes up from one index to the next.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jul 2026

Pinterest MLE interview with a string manipulation problem that sounds deceptively simple until you realize floats will absolutely betray you. The constraint about not using float() or Decimal was the whole point.

Given two numeric strings, one representing a value and one representing a precision, return the value rounded to that precision as a string. For example, rounding '12567' to the nearest '100' gives '12600', or rounding '1234.678' to '0.1' gives '1234.7'. You cannot use float() or Decimal helpers; implement the rounding directly on the string representations.

“My first instinct was to just cast to float and call it a day, and the interviewer shut that down immediately.”

View Post

Pinterest·Machine Learning Engineer·Recruiter / HR Screen

Jul 2026

Interviewed for an ML engineer role at Pinterest, and from what I can tell it was pretty standard screening territory.

Why do you want to work at Pinterest?

“Classic opener.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jul 2026

Pinterest ML Engineer interview with a coding round that leaned more data-structures-heavy than I expected. The main problem was about tracking people moving through numbered rooms and answering count and ranking queries efficiently. Solid problem, took me a while to get the full solution clean.

Design a data structure to track n people moving through n rooms (all starting in room 0). Support two operations: count(roomId) returning the number of people in that room in O(1), and topKFastest(k) returning the k people who have moved the farthest, with ties broken by who arrived in their current room earliest.

“The count part came to me fast, just a counter array per room.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jul 2026

Pinterest MLE coding round, one problem that looks like a classic LC but with a twist that changes everything. Took me a bit too long to realize the left-to-right evaluation rule was the whole point of the problem.

Given a string of digits and a target integer, insert +, -, or * operators between digits (or no operator to form multi-digit numbers) so the resulting expression evaluates to the target. Operators have no precedence and the expression is evaluated strictly left-to-right. Return all valid expressions. Multi-digit numbers cannot have leading zeros.

“I'd seen the LC 282 version before so I jumped straight into a backtracking approach and felt pretty good.”

View Post

Pinterest·Machine Learning Engineer·Onsite - System Design / Architecture

Jul 2026

Pinterest ML engineer interview focused entirely on a large-scale URL health detection system. One long design question that touched basically every layer of the stack, from crawling strategy to downstream ad system integration. Dense but interesting.

Design an ML system to detect broken or low-quality landing pages (404s, soft-404s, parked domains, redirect loops) across a massive URL inventory. Walk through labeling, features, crawling strategy, model architecture, serving, and success metrics.

“This question is basically six questions stitched together and they want you to move through all of them without getting stuck.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jul 2026

Pinterest ML Engineer interview with a tricky warehouse packing problem that adds a twist to a classic LC variant. The bidirectional push mechanic is what makes it interesting and also where I got tripped up.

You have two arrays representing box heights and warehouse room heights. Boxes can be pushed in from either the left or the right side, and each box must fit within the effective minimum height along its entire push path. What is the maximum number of boxes you can fit in the warehouse?

“The bidirectional part is what got me.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jun 2026

Interviewed for an ML engineer role at Pinterest, got a theory question about optimization fundamentals. Pretty standard technical screen vibe, nothing too wild.

What are the differences between convex and nonconvex functions?

“I knew the basics but fumbled a bit explaining why it actually matters for training.”

View Post