Pinterest Machine Learning Engineer (MLE) Interview Guide

Updated 17 days ago by Pinterest candidates

Pinterest machine learning engineer interviewers write their own questions instead of working from a standardized question bank the way Google and Meta interviewers do. Most interviewers at Pinterest focus on the product surface they work on daily, so two candidates for the same role can face entirely different interviews. Preparing for a Pinterest MLE interview means building recommendation systems depth wide enough to cover whichever surface you're handed.

This guide breaks down each stage of the ML engineer interview process, what Pinterest interviewers look for, and how to prepare with real example questions, actionable tips, and resources.

Pinterest machine learning engineer interview process

The Pinterest machine learning engineer interview runs in three stages: a recruiter screen, a technical screen, and an onsite loop of five or more rounds. Expect 3-5 weeks from the recruiter call through the completed onsite, with team matching handled separately once you clear the loop.

Here's an example of what the Pinterest MLE interview process can look like:

1. Recruiter screen: A 30-minute call covering your background, level fit, motivation, and scheduling
2. Technical screen: Roughly 60 minutes combining rapid machine learning questions with one coding challenge
3. Onsite loop: Five or more rounds across coding, ML system design, ML practitioner, and hiring manager behavioral, most running 45 minutes each and system design running 60

Pinterest published a formal policy on AI use in hiring in January 2026. Assume AI assistance isn't permitted on any live interview question unless your recruiter tells you otherwise.

Pinterest interviewers submit a categorical rating after each round rather than a numeric score: strong hire, hire, weak hire, weak no-hire, or strong no-hire. A hiring committee reviews the full set weekly and returns one of three outcomes: hire, no hire, or hire at a lower level.

One no-hire rating is usually decisive. A single weak hire can be offset by strong ratings elsewhere, which is why the ML system design and ML practitioner rounds carry so much of the decision.

Recruiter screen

Expect the Pinterest machine learning engineer recruiter screen to run about 30 minutes and cover your background, your interest in the role, level fit, and scheduling for the rest of the loop. Come in with a compensation range ready, since Pinterest re-leveled its engineering ladder and the band you're quoted maps to a specific IC level.

Most Pinterest recruiter screens stay non-technical, though recruiters occasionally ask a single ML concept question they aren't equipped to evaluate themselves. Give the textbook answer in a sentence or two, then move the conversation forward.

Interviewers look for:

• Level fit: Whether your scope matches the level the team is hiring for
• Motivation: Why Pinterest and why this role specifically
• Machine learning background: The systems you've built and the scale you've built them at
• Compensation alignment: Whether your expectations match the band for the level
• Availability: Your timeline and location for the rest of the loop

Sample questions

Here are some real interview questions reported by candidates:

• Walk me through your resume.
• Why do you want to work at Pinterest?
• What are you looking for in your next role?
• What are your salary and equity expectations?

Technical screen

Pinterest screens ML engineer candidates with a single 60-minute technical call in CoderPad, pairing rapid machine learning questions with one coding challenge. The machine learning portion moves fast and stays on fundamentals like PCA vs. LDA, activation functions, regularization, and the difference between AUROC and AUPRC, so answer in one or two sentences and let the interviewer set the depth.

Expect the coding challenge to run medium to hard difficulty with a follow-up extension built in. Budget your time so the extension has room, since finishing the first part without reaching the follow-up counts as a partial solve.

Interviewers look for:

• Coding fluency: Whether you reach a working solution and extend it when the difficulty increases
• Machine learning recall: Quick, correct answers on core concepts without hedging
• Complexity reasoning: Your ability to state and defend the runtime of what you wrote
• Communication: How clearly you narrate your approach while you write

Sample questions

Here are some real interview questions reported by candidates:

• What's the difference between PCA and LDA?
• How do AUROC and AUPRC differ, and when would you prefer one?
• Where does vanishing gradient show up in a neural network, closer to the input layers or the output layers?
• Which activation function handles vanishing gradient better, ReLU or sigmoid?
• What is regularization, and what are the different types?
• With a million data points, would you use a deep neural network or k-nearest neighbors, and how does inference time compare?
• Given a parent array describing a forest of rooted trees, return the depth of each node.

Coding and data structures rounds

The Pinterest ML engineer onsite includes two coding rounds at mid and senior levels, each running about 45 minutes. Arrays, strings, and hash tables make up the largest share of Pinterest's question set, but graph traversal is emphasized consistently.

Prepare breadth-first and depth-first search first, then arrays, strings, and hashing, then sliding window, backtracking, and occasional class-design prompts that simulate user behavior. Trees appear less often than the graph reputation suggests.

Pinterest frames its coding challenges in product terms, so a graph question might hand you Pin image data as nodes and edges and ask you to surface visually similar images. Restate each prompt as its underlying data structure before you write anything, since the patterns themselves are standard.

Coding carries a recovery path at Pinterest, and interviewers sometimes add a second coding attempt for candidates who come close and perform well in the other rounds. Making no progress at all ends the loop regardless of how strong the rest of your rounds were.

Dynamic programming and backtracking appear in both the technical screen and the onsite, despite an informal internal preference against asking dynamic programming. It shows up often as a follow-up extension on a traversal challenge, so prepare it as a secondary priority behind graph traversal.

Interviewers look for:

• Pattern recognition: Whether you identify the graph or tree structure underneath a product-framed prompt
• Progress under time: How far you get toward a working solution before the round ends
• Follow-up handling: Your ability to extend a first solution when the interviewer raises difficulty
• Trade-off reasoning: How you compare approaches on runtime and memory before committing
• Communication: Whether the interviewer can follow your reasoning as you write

Sample questions

Here are some real interview questions reported by candidates:

• Given a graph of Pin image data, find the images most similar to a target image.
• Given engagement ratings and relevance ratings for a set of Pins, compute a combined ranking.
• Navigate a maze and return the shortest path to the exit.
• Design a class that simulates user behavior on a feed.
• Settle expenses across a group of people where each person owes or is owed a different amount.

ML system design round

Pinterest's ML system design round gives you an open-ended prompt tied to one of its surfaces: related Pins, search ranking, home feed personalization, or ad ranking. Nearly all of them reduce to building a recommendation engine, so the underlying structure repeats no matter which surface you get.

The round runs about 60 minutes, longer than any other round in the loop. Aim to cover the full architecture in the first 20 minutes, which leaves room for at least two areas of depth, either ones you raise yourself or ones the interviewer asks about.

Account for the click feedback loop in your training data and explain why calibration matters for ads before you're asked. Both come up as follow-ups, so prepare a specific example of each from your own work.

Interviewers look for:

• Challenge framing: How you define the label, objective, and success metric before choosing an architecture
• Feedback loop awareness: Whether you account for how click data biases the next training set
• Calibration reasoning: Your understanding of why exact predicted probabilities matter for ads
• Scale judgment: Whether your design holds up at Pinterest's traffic volume
• Trade-off defense: How you justify each component choice when pushed
• Pacing: Whether you finish the architecture with time left for detailed follow-ups

Sample questions

Here are questions shared by a Pinterest interviewer:

• Design a related Pins recommendation model.
• Build Pinterest's search ranking system.
• Design a personalized home feed for Pinterest.
• Build Pinterest's ad ranking system.
• Design a system that recommends visually similar images.
• Design Pinterest's spam detection system.

ML practitioner round

The Pinterest ML practitioner round moves between your past projects and the theory underneath them, which other companies label ML depth, or depth and breadth. Your interviewer reads your resume, picks a project, and works from your decisions into the machine learning concepts behind them.

The round splits into two halves. The first covers one project or paper from your resume in detail, including the challenge, your intuition, the model architecture, and the baselines you compared against.

The second half moves into theory. Expect questions on transformer architecture, contrastive loss, and learning to rank, alongside tuning judgment: when to change batch size, how to set learning rate decay, what shifts when you swap optimizers, and what to do when training loss diverges.

Interviewers look for:

• Diagnostic sequencing: Whether you locate the bottleneck before you choose a fix
• Tuning judgment: Your reasoning for changing one hyperparameter before another
• Project ownership: How much of the modeling work on your resume you drove yourself
• Failure recall: What went wrong in your models and how you found it
• Theoretical grounding: Whether you can explain the concepts behind the choices you made

Sample questions

Here are questions shared by a Pinterest interviewer:

• Walk me through a machine learning project you owned end to end, including your baselines.
• Would you increase or decrease the learning rate if your loss is oscillating?
• Explain how a transformer architecture works.
• What is contrastive loss, and when would you use it?
• When would you change a model's batch size, and what breaks when you do?
• How do you set learning rate decay?
• What shifts when you swap the optimizer?
• Your model's accuracy is poor. How do you find out why?
• Explain how a two-tower architecture works and how you'd tune one.
• Explain data drift and how you'd detect it in production.

Hiring manager behavioral round

Pinterest's hiring manager round, which candidates also call the competency round, is the only real behavioral evaluation in the machine learning engineer loop. It runs as an open conversation without a named leadership framework behind it, and your interviewer submits the same categorical hire rating used in every other round.

Hiring managers are checking whether your scope matches the level, how you work with partners outside your own team, and how you respond when someone challenges your approach.

Pinterest publishes five company values: Put Pinners First, Aim for Extraordinary, Create Belonging, Act as One, and Win or Learn. Read them for context on the culture, and build your examples around the work itself.

Interviewers look for:

• Level fit: Whether your scope and judgment match the level you're interviewing for
• Stakeholder communication: How you translate modeling trade-offs for product and engineering partners
• Execution: Whether you shipped what you set out to ship and can name what slipped
• Openness to feedback: What you changed after a review challenged your approach
• Collaboration: Whether you help teammates succeed and share credit for outcomes

Sample questions

Here are some real interview questions reported by candidates:

• Tell me about a time you received critical feedback. What did you change?
• Describe a disagreement with a teammate and how you resolved it.
• Walk me through a project where you worked across several teams.
• Tell me about someone you mentored and what changed for them.
• Give me an example of a technically difficult decision you made and had to defend.

How to prepare for the Pinterest machine learning engineer interview

1. Confirm which product surface your team owns: Ask your recruiter whether the team works on home feed, search, ads, or shopping before you start preparing. Pinterest interviewers build questions around the systems they work on daily, so the surface tells you which recommendation architecture to prepare first.
2. Read Pinterest's engineering blog for the surface you're interviewing for: Work through recent posts on home feed ranking, search, ads retrieval, and shopping conversion, noting the model names, training signals, and serving constraints in each. Pinterest's 2026 work on foundation model training scalability, generative retrieval, request-level deduplication, multi-objective home feed optimization, and its move beyond two-tower serving for ads gives you specific systems to reference in the design round.
3. Prioritize graph traversal: Practice breadth-first search, depth-first search, and shortest path until you can write them without reference, then cover arrays, strings, and hash tables, then sliding window, backtracking, and dynamic programming. Train yourself to restate a product-framed prompt as its underlying data structure before writing anything, since Pinterest wraps standard patterns in Pin and user data.
4. Practice recommendation system design on a 20-minute timer: Cover requirements, candidate generation, features, model choice, training data, serving, and evaluation end to end within the limit, then use the remaining time for two areas of depth. Repeat the exercise across related Pins, search ranking, home feed, and ad ranking until the structure comes automatically.
5. Prepare a critique of your own design: Name the weakest component, the traffic volume at which it breaks, and the fix you'd ship first. Define your latency, throughput, and freshness requirements before you name a specific database or message queue.
6. Prepare two model failure stories: For each one, describe the metric that dropped, how you isolated the cause, and the order you changed hyperparameters in. Pick one case where the data caused the issue and one where the architecture did.
7. Practice in both AI-permitted and unaided modes: For rounds where AI assistance is allowed, practice narrating your prompts and how you verify the output. For the rest, write and explain your code with autocomplete disabled.
8. Prepare for LLM and generative AI questions: Pinterest now folds questions about large language models into its machine learning rounds and lists experience with AI coding assistants as a preferred qualification in its 2026 postings. Be ready to discuss where you've applied generative models and which coding assistants you use in your own workflow.
9. Book a mock interview: Run a full ML system design mock interview with a peer to get comfortable holding a 20-minute pace under pressure. Follow it with expert coaching from an ML engineer who can push your architecture and tuning answers the way a Pinterest interviewer would.

About the Pinterest machine learning engineer role

Pinterest machine learning engineers build and ship the ranking and retrieval systems behind the products Pinners use daily, from home feed personalization to search to ads. Most work sits inside a recommendation pipeline, covering candidate generation, feature engineering, model training, and serving at production scale.

Pinterest renumbered its engineering ladder to IC13 through IC18, where IC15 is Senior Software Engineer and IC16 is Staff. Machine learning engineers share that ladder with software engineers as a job-title variant in the same role family, so your IC level sets both your offer band and your interview bar. Open roles and reported compensation both skew senior.

Pinterest staffs MLEs across these teams and surfaces:

• Homefeed personalization and related Pins
• Search and visual search ranking
• Ads and Monetization, covering ads ranking, ads ML infrastructure, and vertical modeling
• Content Shopping and shopping conversion
• Growth, User Understanding, and Curation ML
• Core Engineering, covering embedding and retrieval infrastructure
• Pinterest Labs, covering applied research and foundation models

Pinterest restructured in January 2026, cutting under 15% of its workforce while shifting resources toward AI roles. Machine learning hiring continues, and team matching can take longer than the interview loop itself.

New engineers go through Basecamp, Pinterest's engineering onboarding program, with an assigned buddy for the first several weeks. Expect meaningful output around month three or four, and note that Pinterest sets goals at the team level on quarterly, half-year, and annual horizons. Ask about the ramp plan in your hiring manager round to learn how much support the team gives new hires.

Pinterest machine learning engineer experience requirements

Pinterest states required experience by level in its ML engineer job postings:

Every level asks for a degree in computer science, machine learning, or statistics, with a master's or PhD preferred at the II level. Postings consistently name Python and Java, PyTorch or TensorFlow, and Spark or Hadoop, alongside experience in recommendations, ranking, search, natural language processing, reinforcement learning, or graph representation learning.

Pinterest's 2026 postings also list familiarity with AI coding assistants and large language model productivity tools as a preferred qualification. Name the tools you use and how you verify their output.

How the Pinterest MLE loop changes at staff level

Staff-level candidates typically get a single coding round in place of two, at least two behavioral rounds including one with a manager from another team, and often a conversation with a VP. Some senior candidates at IC15 may also see two coding rounds, so confirm your loop composition with your recruiter rather than assuming the reduction applies at your level.

Prepare for the behavioral rounds to carry more of the decision at these levels too, with cross-team scope and technical direction as the recurring themes.

FAQs about the Pinterest machine learning engineer interview

How long is the Pinterest machine learning engineer interview process?

The Pinterest machine learning engineer interview process typically takes 3-5 weeks from the recruiter call through the completed onsite, with an average around 25 days. Post-onsite feedback runs 1-2 weeks, and rejections tend to arrive faster than offers. Team matching happens after the loop, draws from a shared candidate pool, and has carried a backlog since Pinterest cut under 15% of its workforce in January 2026, with some candidates waiting a month or more. Passing the loop doesn't guarantee placement, so ask your recruiter how many teams are actively matching before you plan around competing offers.

How much does a Pinterest machine learning engineer make?

Pinterest machine learning engineers sit on the same IC13 through IC18 ladder as software engineers, so compensation reports carry software engineering titles. Here are the reported compensation ranges by level for Pinterest machine learning engineers, according to Levels.fyi:

• IC13 (Software Engineer I): ~$229K
• IC14 (Software Engineer II): ~$308K
• IC15 (Senior Software Engineer): ~$482K
• IC16 (Staff Software Engineer): ~$704K
• IC17 (one level above Staff): ~$1.04M

The reported median across all levels is roughly $392K. Packages combine base salary, bonus, and RSUs vesting over 4 years. Pinterest reduced its compensation bands during the IC13 to IC18 re-leveling, so figures published before 2025 overstate current offers.

How many rounds is the Pinterest machine learning engineer interview?

The Pinterest machine learning engineer interview runs five or more rounds across three stages. Expect a 30-minute recruiter screen, a 60-minute technical screen, and an onsite loop of two coding rounds, one ML system design round, one ML practitioner round, and one hiring manager behavioral round. Roles with an infrastructure focus sometimes add a sixth onsite round, and staff-level candidates trade one coding round for additional manager conversations.

Does Pinterest allow AI assistance during interviews?

Pinterest prohibits AI assistance on live interview questions by default. The company's January 2026 hiring policy directs candidates to assume AI isn't permitted unless a recruiter says otherwise, and Pinterest also runs designated segments where AI use is allowed and evaluated. Ask your recruiter which rounds fall into which category before your loop begins.

Does Pinterest ask dynamic programming questions?

Dynamic programming does come up in the Pinterest machine learning engineer interview, in both the technical screen and the onsite. Pinterest discourages dynamic programming questions internally, but interviewers choose their own material, so that guidance isn't consistently followed. Dynamic programming appears in roughly one in six of Pinterest's questions, often as a follow-up extension rather than the main challenge. Prepare it as a secondary priority behind graph traversal, arrays, and hashing.

How is the Pinterest ML engineer interview different from Google or Meta?

The Pinterest machine learning engineer interview differs from Google and Meta in three ways. Pinterest runs a dedicated ML practitioner round that examines your own projects and tuning decisions, which Google and Meta fold into broader ML depth rounds. Pinterest evaluates behavior only in the hiring manager round, with no named leadership framework scoring your answers. Pinterest also lets interviewers write their own questions, so your loop reflects the product surface your interviewers work on more than any standardized company rubric.

What's the difference between the ML system design and ML practitioner rounds at Pinterest?

Pinterest's ML system design round evaluates whether you can architect a recommendation system end to end at Pinterest's traffic volume, covering candidate generation, features, model choice, serving, and evaluation. The ML practitioner round covers your own past projects and the modeling decisions inside them, including how you tuned hyperparameters and diagnosed failures. Prepare design by practicing full architectures on a timer across related Pins, search, home feed, and ads. Prepare for the practitioner round by writing out two of your own projects in detail, including the metrics that dropped and the order you changed hyperparameters in.

Can you fail the Pinterest coding round and still get an offer?

You can recover from a weak coding round at Pinterest if you reach a near-working solution and perform well in the ML system design and ML practitioner rounds, since interviewers sometimes add another coding round for candidates in that position. A weak hire rating can be offset by strong ratings elsewhere, but a no-hire rating on any round is usually decisive. Making no progress on the challenge ends the process regardless of how the rest of the loop went. Protect against that by writing a working brute-force solution first, then optimizing from there, so you always have something functional to discuss. State your approach and its complexity out loud before you write, which gives the interviewer a correct plan to credit even if you run out of time on the implementation.