Alien Dictionary (LeetCode 269): Given words sorted in an unknown alien language's lexicographic order, derive the ordering of letters in that language.

The key insight is that comparing adjacent words character-by-character reveals ordering edges between letters, which you then topologically sort (BFS/Kahn's or DFS with cycle detection). Build a directed graph of char constraints, run topological sort, detect cycles (invalid input). Time O(C) where C is total characters. Justin Holiday reported that three different Pinterest interviewers asked this exact problem in his 2020 loop, making it one of the most reliably confirmed Pinterest questions. Ability to model a constraint problem as a directed graph and apply topological sort; handling edge cases like prefix words in wrong order (invalid) and characters with no constraints.

K-th Biggest Pair Sum: Given two sorted arrays, find the k-th largest sum formed by picking one element from each array.

Use a max-heap seeded with the largest pair (A[last], B[last]) and expand by decrementing indices, using a visited set to avoid duplicates. Pop k times. Time O(k log k). Justin Holiday reported this as the exact phone-screen question asked by a Pinterest recruiter for a Toronto SWE role in late 2020. Heap-based k-th order statistics reasoning, efficient state expansion, and handling of duplicate index pairs cleanly.

Binary Tree Zigzag Level Order Traversal (LeetCode 103): Return the nodes of a binary tree in zigzag level order as a 2D array, alternating left-to-right and right-to-left by level.

BFS with a deque; alternate appending to front or back based on level parity. O(n) time and space. Multiple Glassdoor reviewers (QTN_1669937) explicitly report this as a Pinterest onsite coding question alongside LRU Cache. The follow-up is often to extend to an N-ary tree. Clean BFS implementation, correct alternation logic, and ability to extend the pattern to related traversal variants.

Design a system where users can follow Pins or other users and receive notifications when the followed entity is updated.

Justin Holiday documented this as his exact 2020 Pinterest onsite system design question. Core components: a follows table (user_id, entity_id, entity_type, created_at) in a relational DB; a fan-out service triggered by pin update events (Kafka topic); notification delivery via push (APNs/FCM) and in-app feed backed by Redis. Discuss write-time vs read-time fan-out trade-offs given Pinterest's follower graph skew. Follow-up is usually around handling unfollow/notification suppression and deduplication. Data modeling for a polymorphic follow relationship, event-driven fan-out architecture, and notification delivery at scale.

Design a system to detect visually similar products across Pinterest's Pin catalog and surface them to users (Similar Products feature).

Justin Holiday documented this as his 2021 extended Pinterest system design interview. Pipeline: image embedding service (CNN-based) runs offline over all Pins to produce feature vectors stored in a vector index (FAISS/ANN); at query time, embed the query Pin and nearest-neighbor search the index. Online ranking layer re-ranks by engagement signals. Discuss index freshness (incremental updates), latency SLOs, and cold-start for newly uploaded Pins. ML-powered retrieval pipeline design, vector indexing trade-offs, and separation of offline embedding generation from online serving.

Design the Pinterest home feed: a personalized, infinite-scroll feed of Pins for a logged-in user.

This is the canonical Pinterest system design prompt confirmed across multiple candidate guides and Glassdoor SWE reviews. Hybrid push-pull fanout: pre-compute feeds for users with few following relationships (push model via Kafka + Redis sorted set); fetch-on-read for high-follower accounts (pull model). Ranking service applies a learned model over candidate Pins. Discuss pagination cursor design for infinite scroll, cache invalidation, and cold-start for new users. Feed generation at scale, push vs pull trade-off reasoning tied to follower-graph distribution, and ranking pipeline design.

Tell me about a time you disagreed with a teammate or engineering manager on a technical decision. How did you resolve it?

Pinterest explicitly looks for data-driven, collaborative conflict resolution rather than authority or seniority-based deference. Structure the answer with STAR: the technical disagreement, the data or prototype you used to make the case, how you listened to the counterargument, and the agreed-upon outcome. Pinterest's first value is 'put pinners first' so tying the resolution to user or product impact lands well. Constructive disagreement, use of evidence over opinion, respect for peers, and ultimate alignment around user or product outcomes.

Describe a project where you had to work across multiple teams or functions (e.g., with design, product, data science). What was your role and what made it succeed or fail?

Pinterest emphasizes cross-functional collaboration ('Knit' as a cultural value). Highlight your specific technical contributions, how you unblocked others, and any communication or alignment challenges. Be concrete about the outcome and what you would do differently. Avoid vague team-speak; interviewers want to know what you personally owned. Ability to drive technical work in a multi-stakeholder environment, ownership clarity, and honest reflection on what succeeded and what did not.

Tell me about a time you made a significant technical mistake or shipped something that caused a production incident. What did you do?

Pinterest interviewers use this to gauge ownership, humility, and learning. Lead with what went wrong, your role in it, how you detected and mitigated the impact, what the root cause was, and the concrete process or code changes made afterward. Avoid blaming others. Strong answers include a monitoring or testing change that prevents recurrence. Personal accountability, systematic root-cause thinking, and demonstrated learning that changed behavior.

Pinterest Interview Questions

Pinterest runs a 4-5 round interview loop for mid-to-senior SWEs (L4/L5): a CodeSignal or CoderPad phone screen, two coding rounds, one system design round, and a behavioral/hiring-manager round. The loop is conducted over one to two virtual days with 45-minute sessions each. Pinterest allows use of online resources like Google and Stack Overflow during coding rounds to simulate real work, and steps problems incrementally to gauge depth. The company added a CodeSignal OA as an early screening gate for new-grad pipelines starting around 2022.

17 real questions across 3 rounds

Reported via Justin Holiday first-hand Medium posts (2020 Pinterest phone interview: k-th biggest pair sum; 2020 Pinterest algorithm interview: Alien Dictionary; intuting.medium.com), LeetCode Discuss thread by candidate reporting graph path/DFS phone screen (leetcode.com/discuss/interview-question/906065), Glassdoor SWE reviews reporting zigzag binary tree + LRU cache + autocomplete multi-step questions (glassdoor.com, QTN_1669937), Glassdoor February 2024 report of LeetCode 1564 (Put Boxes Into the Warehouse I) with follow-up, AlgoDaily/Prepfully aggregating Glassdoor-sourced Pinterest questions (algodaily.com/companies/pinterest), Pinterest Engineering Blog confirming multi-step incrementally harder format (medium.com/pinterest-engineering)

Run a Pinterest mock with the AI interviewer. It asks questions like these, follows up on your answers, and tells you exactly what to fix. The only score that means anything is the one on your real answers.

Continue with Google

Keep prepping

All company interviews