← Back to Directory

Pinterest interview notes, page 20 of 24

Large Enterprises

Most of these Pinterest write-ups come from Machine Learning Engineers (91), then Software Engineers (77) and Data Scientists (51). The most reported rounds are the technical phone screen (168), the system design round (37), the product sense round (14) and the behavioral / leadership round (5).

235 interview notes · updated Sep 2026

Pinterest·Software Engineer·Technical Phone Screen

Jun 2026
Pinterest coding screen, one question, combinatorics flavored. Pretty straightforward but the kind of thing you want to have seen before so you don't fumble the implementation.
  • Given a dictionary where each key maps to a list of possible values, generate all combinations as a list of dictionaries, one per combination.

“Classic cartesian product problem.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jun 2026
Pinterest DS interview focused entirely on one big product case: a carousel feature for the Home feed. The whole thing was a deep dive into metrics, experimentation design, and diagnostic thinking. Pretty intense for a single question.
  • Before evaluating a new carousel feature on the Home feed, what product goals would you clarify first, and how would those goals shape your evaluation framework?
  • Define primary, secondary, and guardrail metrics for the carousel. How do you think about trade-offs between module CTR, overall Home CTR, saves, session depth, retention, and monetization impact?
  • What events, dimensions, and logging would you instrument to support analysis of this carousel feature?
  • How would you design an A/B test or staged rollout for this feature? Cover randomization unit, exposure definition, novelty effects, power calculations, and potential sources of bias.
  • The feature launches and overall Home-page CTR drops. Walk through the most plausible explanations and how you'd distinguish between a logging bug, a denominator shift, feed cannibalization, segment-specific effects, and a genuine product regression.

“I started by separating engagement goals from monetization goals because those pull the framework in totally different directions.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Pinterest MLE coding round, two problems back to back. One was genuinely hard, the other felt like a warmup they threw in at the end.
  • Given an array of bar heights representing a histogram where each bar has width 1, find the area of the largest rectangle that can be formed within it.
  • Given two strings s and t, determine whether s is a subsequence of t, meaning all characters of s appear in t in the same order but not necessarily consecutively.

“This is the kind of problem where you either know the monotonic stack trick or you're cooked.”

View Post

Pinterest·Product Manager·Onsite - Product Sense / Strategy

Jun 2026
PM interview at Pinterest with a product redesign question about breaking into a new vertical. Just the one question but it took up the whole session, which I wasn't expecting.
  • If you were a PM at a major workplace collaboration tool, how would you redesign it to enter the education sector?

“I spent too long on the feature brainstorm and not enough time defining who the actual user was.”

View Post

Pinterest·Software Engineer·Technical Phone Screen

Jun 2026
Pinterest coding interview, algorithm-heavy with a grid-based robot vacuum problem that had multiple variants. The question escalated fast and I wasn't fully prepared for the coordination part.
  • Given a 2D grid with obstacles and two robot vacuums each starting at a different cell, how would you compute the union of all cells reachable by either robot? Each robot can move in 8 directions.
  • Now coordinate the two robots so they together clean every reachable cell, optimizing for either minimum total moves or minimum makespan (the longer of the two individual paths). How does your approach change based on which cost metric you pick?

“Part (a) felt manageable.”

View Post

Pinterest·Machine Learning Engineer·Online Assessment (OA)

Jun 2026
Pinterest ML engineer interview had a multiple-choice section where you actually had to do neural network forward pass math by hand. Nothing crazy, but it caught me a little off guard that they expected you to crunch numbers rather than just talk through concepts.
  • Given small weight matrices and a bias vector, manually compute the output of a fully-connected layer followed by one or two activation functions (e.g. ReLU, sigmoid, softmax) applied end-to-end.

“The arithmetic is kept simple on purpose, which almost made me second-guess myself.”

View Post

Pinterest·Data Analyst·Technical Phone Screen

Jun 2026
Pinterest health-tech product analytics interview, one SQL question that looked clean on the surface but had a few layers to it once you got into the tie-breaking and the 'strictly after' condition. Felt like a take-home style problem delivered live.
  • Given an appointments table with statuses like cancelled, completed, no-show, etc., write a SQL query to find what percentage of users whose first appointment was cancelled never went on to book another returned or confirmed appointment afterward.

“The tie-breaking rule tripped me up a bit.”

View Post

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

Jun 2026
Pinterest ML Engineer system design round, focused entirely on embedding infrastructure at scale. The question was meaty and covered a lot of ground, from training paradigms to cold-start handling to ANN indexing. Left feeling like I did okay on the core architecture but probably undersold the evaluation side.
  • Design a system that learns and serves embeddings for users and items, to be used as input for downstream personalization tasks like recommendations, ranking, search, and ads.

“This one sprawls fast if you let it.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jun 2026
Pinterest data science coding screen, one question the whole time. Pretty focused on correctness plus efficiency, and they wanted unit tests at the end which I almost forgot about.
  • Write a function that takes a mapping of users to lists of pin IDs and returns the mean number of unique pin IDs per user, rounded to two decimal places. Empty or missing lists count as zero, None values inside lists should be ignored, and the solution should run in O(total items) time with O(unique pins) space.

“The deduplication part was fine but I initially used a global set across all users which tanked my space complexity.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jun 2026
Pinterest DS interview with a heavy quantitative case study involving out-of-home media measurement. The whole thing was basically one giant problem broken into three parts, and it got pretty deep into probability, Markov chains, and bias analysis. Felt more like a stats exam than a typical product DS interview.
  • Given traffic counts, vehicle occupancy, visibility probabilities by time of day, and pass frequency distributions by user segment, estimate weekly total impressions and unique reach for a digital billboard, including 95% confidence intervals. Walk through your formulas and state your independence assumptions explicitly.
  • Model the path from billboard exposure to store visit using a 3-state Markov chain (Unaware, Aware, Visit). Propose transition probabilities by user segment, compute expected weekly visits attributable to the billboard, and discuss how sensitive the results are to your chosen probabilities and to the visibility parameter.
  • Identify at least three significant sources of bias in this measurement setup and propose concrete corrections or validation approaches for each.

“This is the part I felt most prepared for and still got tripped up.”

View Post