Pinterest Interview Questions — Page 4

Pinterest·Data Scientist·Technical Phone Screen

Apr 2026

Pinterest DS interview that was basically a stats exam disguised as a conversation. Heavy on ads measurement theory and then they made you grind through the actual math by hand, which I was not fully prepared for.

How would you define Brand Lift Study versus Conversion Lift Study in ads measurement, and what are the main sources of bias and variance in each?
When would a Brand Lift Study and a Conversion Lift Study give you conflicting results, and how would you go about reconciling them?
Compare Difference-in-Differences and Propensity Score Matching for observational ads data. What are the identification assumptions for each, and what robustness check would you run?
In a randomized conversion lift holdout, treatment group had 1200 users with 42 conversions and control had 1180 users with 33 conversions. Compute the point lift, the pooled standard error under the null, and the two-sided z-statistic. Is the result significant at the 5% level?
You have pre/post Purchase Intent survey data for exposed and control groups with given sample means and variances. Derive the DiD estimator, compute an approximate standard error using the four independent variance terms scaled by their sample sizes, and report the t-statistic. Show your formulas.

“I knew the high-level distinction but stumbled when they pushed on specific bias sources.”

View Post

Pinterest·Software Engineer·Technical Phone Screen

Apr 2026

Pinterest SWE interview with a grid simulation problem involving lighthouses, mirrors, and walls. Pretty niche problem type, not your typical graph traversal, and the cycle detection piece is where things get interesting.

Given a 2D grid containing lighthouses, walls, mirrors, and empty cells, determine how many cells are illuminated. Lighthouses emit rays in all four cardinal directions, mirrors reflect rays 90 degrees, walls block rays, and a cell is illuminated if any ray passes through it.

“The mirror reflection logic tripped me up more than I expected.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Apr 2026

Pinterest data science interview with a system design question that leaned way more engineering than I expected for the role. The problem was interesting but felt misaligned with what a DS actually does day to day.

Design and implement a permission system that supports geographic path-based access control (e.g., /France/Paris). The system needs to support adding a path and checking whether a given path is authorized.

“I went straight to a hash map and the interviewer kind of waited, like they wanted something more structured.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Apr 2026

Pinterest data scientist interview that went deep into survey methodology and probability. The whole session was basically one long scenario about a gender-imbalanced user survey, and they kept pushing on the stats fundamentals underneath it.

You ran a user satisfaction survey and noticed the gender split in your respondents doesn't match the actual user population. What are the key concerns with using this data to estimate overall satisfaction?
How would you validate whether the gender imbalance actually biases your satisfaction estimate? What checks or external data would you use?
What distributional assumptions would you make for the gender counts in your sample and for the satisfaction outcome within each gender? When are those assumptions reasonable?
Compute the probability of observing exactly 30 female respondents out of 70 total, first assuming independent sampling from a population with true female proportion p, then assuming sampling without replacement from a finite population of size N with F females.
How would stratified sampling help with the gender imbalance problem, and if you had used it, how would you combine results across strata to estimate overall satisfaction?

“Felt pretty comfortable here.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

Apr 2026

Pinterest ML Engineer interview had a coding round focused on implementing a bagging classifier from scratch, no NumPy allowed. Pretty implementation-heavy, less theory than I expected.

Implement a bootstrapping function that samples n training examples with replacement from a dataset, returning the sampled features and labels.
Implement the fit method for a BaggingClassifier that trains multiple decision trees, each on a different bootstrap sample of the training data.
Implement the predict method using majority vote aggregation across all trained trees, and explain how you handle tie-breaking.

“Straightforward once you realize it's just random.choices with k=len(X).”

View Post

Pinterest·Software Engineer·Technical Phone Screen

May 2026

Pinterest phone screen for a software engineering role, one coding problem the whole time. Pretty focused on data structures and complexity analysis, not just getting the code to work.

You're given a call log of phone numbers and a list of user-reported spam numbers (with possible duplicates). Cross-reference the two to output each spam number that appears in the call log along with how many times it was reported. Walk through your algorithm, choice of data structures, how you'd handle invalid numbers, and the time and space complexity. Then write the code.

“The core idea clicked fast, use a hash map to count reports, then check each call log entry against it.”

View Post

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

May 2026

Pinterest ML engineer interview focused entirely on designing a conversational recommendation bot for Pins. It was a deep system design session covering a lot of ground, from retrieval to ranking to latency constraints, and I felt like I was playing catch-up the whole time.

Design a personalized conversational bot for Pinterest that takes a user's chat input and history and recommends relevant Pins. Walk through the full system: intent understanding, candidate retrieval, ranking, session management, latency, evaluation, and feedback loops.
How would you handle cold-start for new users in this conversational recommendation system?
How would you evaluate this system both online and offline, and what feedback loops would you build in?
What are the main safety and content freshness concerns in a Pin recommendation system, and how would you address them?
How would you design the system to meet the latency requirements of a real-time chat experience?

“This is one of those questions where the scope is so big you can either go broad and shallow or pick a thread and pull it.”

View Post

Pinterest·Machine Learning Engineer·Technical Phone Screen

May 2026

Pinterest ML Engineer interview that came down to implementing a Naive Bayes classifier from scratch. Pretty focused session, no fluff, they wanted to see if you actually understood the math behind something you've probably just called from sklearn a hundred times.

Implement a Naive Bayes classifier from scratch. It should handle both Gaussian (continuous features) and Multinomial (word-count features) variants, support a fit method that computes class priors and per-class likelihoods, and a predict method that returns the argmax of the log-posterior across classes. Use log-space arithmetic and Laplace smoothing where needed.

“First thing they asked was which variant I wanted to implement and I kind of fumbled that.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

May 2026

Pinterest data science interview that was heavier on coding fundamentals than I expected. The main problem was a combinatorics/enumeration thing and then they wanted a full test suite on top of it, which I wasn't totally prepared for.

Given several groups of hyperparameters where each group is a list of allowed values (e.g., learning rate, batch size, optimizer), write a function that returns all possible combinations across all groups, i.e., the Cartesian product.
Design test cases for the hyperparameter combination function, covering: empty input, a single group with one value, a single group with many values, multiple groups, duplicate values within a group, and whether the output ordering is consistent across runs.

“I knew this was itertools.product territory immediately but fumbled explaining WHY that was the right abstraction.”

View Post

Pinterest·Software Engineer·Technical Phone Screen

May 2026

Pinterest iOS engineering interview that was basically one big design problem: build a flood-fill app from scratch and talk through every layer of it. Covered UI architecture, data modeling, algorithm choices, and edge cases all in one shot. Felt more like a system design round than a coding screen.

Design and implement an iOS app that shows a 2D grid of colored cells. Tapping a cell should flood-fill all 4-directionally adjacent cells of the same color to the tapped cell's color.
What data structure would you use to represent the grid, and why?
Walk through the flood-fill algorithm. Would you use iterative or recursive, and what are the time and space complexity trade-offs?
How would you handle very large grids efficiently and avoid stack overflow?
How would you implement reset and undo functionality for the grid?

“This was the whole interview basically.”

View Post