Pinterest Machine Learning Engineer (MLE) Interview Guide

The gist

Pinterest is one of the world’s most popular social media and photo-sharing sites, serving 500 million users a month. Due to its unique interface, in which users create “boards” for a particular theme and “Pin” photos, links, and videos that fit the category, Pinterest also behaves as a semantic search engine across many types of online media. A user who searches for a term like “cyberpunk fashion,” for example, will discover boards full of Pinned links to matching photoshoots, outfits, and digital storefronts picked out by other users.

Pinterest’s headcount is fairly small for a company operating one of the internet’s most popular websites, at roughly 4000 employees. Engineers at Pinterest have more responsibilities, product ownership, and flexibility expected from them than their peers at larger tech companies like Meta or Google.

Working across many different tech stacks is common. While this may sound daunting, Pinterest cultivates a friendly and open office culture that encourages employees to collaborate and share their knowledge with each other. New hires are often paired with an onboarding buddy to show them the ropes for the first month or two. Pinployees praise the company’s welcoming atmosphere and generous perks, such as their versatile PinFlex remote work policy.

What does a Pinterest Machine Learning Engineer do?

ML engineers at Pinterest have a hand in almost every aspect of the company’s business model. They build the systems that power Pinterest’s search algorithm, recommendation engine, personalized user feeds, ad delivery, user acquisition, and more. Pinterest team members are expected to work on cross-functional teams where they may collaborate with stakeholders from several other departments.

Most Pinterest ML engineers work with a tech stack based on Python and PyTorch, and use data processing tools like SQL, Spark, and Hadoop. Their job can involve a wide variety of ML methods and approaches, such as user modeling, personalization, recommender systems, search, ranking, natural language processing, reinforcement learning, and graph representation learning. Due to Pinterest’s massive user base and reach, much of their infrastructure is distributed and based on containerization tools like Docker and Kubernetes.

While ML is invaluable to Pinterest’s bottom line, the company is also dedicated to using it in responsible and inclusive ways. Pinterest ML Engineers can expect to work on—or with—teams that identify, avoid, and mitigate potential instances of bias or harm that may result from certain AI-powered applications.

The average total compensation across software engineering levels at Pinterest are:

• (Entry-level) Software Engineer I: $228k
• Software Engineer II: $354k
• Senior Software Engineer: $469k
• Principal Software Engineer: $714k

Before you apply

• Brush up on your knowledge of machine learning algorithms and systems, including how they perform in real-world situations.
• Take a day or two to read Pinterest’s engineering blog, especially posts about how Pinterest engineers designed their systems and especially those relating to the team or product you’re applying to work on. The most heavily weighted rounds of the Pinterview are the system design questions, and they usually focus specifically on the system you’ll be assigned to. Pinterest’s tech blog will give you a deeper understanding of how these systems are built, and the articles themselves are good examples of how to walk an unfamiliar person through the steps and considerations of a system design problem.
• Research the recent interview questions asked at Pinterest.

Interview process

Applicants to Pinterest complete a sequence of three steps before they get hired:

1. A recruiter phone screen to check that you meet the minimum role requirements.
2. A technical screen conducted over phone or video chat, focused on conceptual ML questions and a coding problem or two.
3. A final interview loop of 5–7 rounds, with 5 being the most common. This is usually held on-site, but may be conducted remotely.

1. Recruiter interview

A Pinterest recruiter or HR rep will reach out to fill you in on what the role entails, and to ask about your experience and your interest in the job. Usually they’ll schedule a short (~30 minute) phone call, although in some cases this whole exchange can take place over e-mail. The recruiter could ask a simple technical question or two, but only rarely.

Primarily this round will focus on your previous work history, career goals, and scheduling the remaining steps of your Pinterview.

Prepare for questions like:

• 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?

2. Technical screen

This round will last about 45–70 minutes, and is typically conducted through a program like CoderPad or LeetCode over a call with a Pinterest Engineer. Most candidates report answering between 3 and 5 ML questions, with a few minutes for each, and then a single medium-hard coding challenge. (It will often be split into several objectives of increasing difficulty.) You may get two easier coding problems instead, or even three (skipping the knowledge questions), but this is uncommon.

Sample questions include:

• Does the vanishing gradient problem appear closer to the beginning or end of a neural network?
• Which activation function is better at mitigating the vanishing gradient problem, ReLU or sigmoid?
• What is regularization? What are the different types of it?
• If you have a million data points, would you use DNN or KNN? What's the difference in inference time?

Sample coding problems include:

• Write a program that finds the best way to settle expenses between a group of friends, each of whom owes or is owed a different amount.
• Implement the K-nearest neighbors algorithm.

3. Final interview rounds

The final piece of the Pinterview is several hours long, and takes place on-site or over a video call. It can be scheduled for a single day (with a lunch break halfway through) or with each half split across two days. This stage is divided into 5+ interview sessions, each focused on a different subject area.

Expect:

• Two rounds focused on solving a coding problem with data structures and algorithms.
• One round focused on conceptual ML questions and your past ML experience.
• At least one round focused on ML system design. There may be two, especially if you’re interviewing for a more senior role or a particularly infrastructure-focused one.
• One round of behavioral questions with the role’s hiring manager. If you’re applying for a role at Staff level or above, you may get two or even three behavioral interviews (replacing one of the coding rounds). The extra rounds will be conducted by a VP or manager from a different Pinterest team.

Data structures & algos coding rounds

A coding round at Pinterest is usually modeled after a LeetCode-style data structures problem, but framed in terms that are relevant to Pinterest product engineering. For example, you might be presented with a graph full of Pinterest image data and asked to write a program that identifies similar images.

Topics that are likely to be focused on include:

• Graphs
• Trees
• Sliding Window
• Backtracking
• Matrix operations

Much like the first technical screen, these coding problems are also likely to present several tasks of increasing difficulty, and there will often be several possible approaches to solve them. Be sure to communicate your thought process to the interviewer, ask clarifying questions, and think through the problem.

ML concepts and practice

After the ML system design rounds, this section holds the next greatest weight in Pinterest’s hiring decision. The interviewer will dive into your past projects and test your knowledge of specialized ML topics. This won’t necessarily be a rapid-fire Q&A session but rather a discussion of past and current projects, where the interviewer segues into technical questions at certain points to evaluate the limits of your knowledge in a particular area.

Many senior Pinterest engineers (including, most likely, your interviewer) work on one product area for a long time, and acquire deep familiarity with it. They’ll ask questions about concrete performance concerns as much or more than abstract ML concepts.

Sample questions include:

• Tell me about a machine learning project you worked on.
• When should you change an ML model’s batch size?
• How do you set an ML model’s learning rate decay?
• What happens when you change an ML model’s optimizer?
• Explain overfitting.
• Explain regularization. What are the different types of it?
• Explain the differences between convex and non-convex functions.
• Explain the differences between ReLU and sigmoid.
• For a data set with a million data points, would you use DNN or KNN? Why?
• Explain data drifting.

ML system design

This round (or rounds!) will have the greatest impact on whether you get the job (and level) you desire. You’ll be given an open-ended problem and ~30 minutes to describe how you’d build a technical solution for it. Most often, the problem will relate to a specific Pinterest product or business need, such as ranking ads or personalizing a user’s home feed. On the plus side, many of the possibilities share fundamental aspects in common—for example, ads, search, and personalization systems are all variations on a recommendation engine.

A good system design answer starts by thinking through the problem, contextualizing where it fits in the Pinterest ecosystem, and foreseeing the pros and cons of a particular approach. In general, as you lay out your design, you should briefly address the trade-offs you’re making as you go, but stay focused on presenting the overall picture. You can (and should) address the details in more depth afterward, or when the interviewer asks follow-up questions.

Sample questions include:

• Design a product recommendation system.
• Design Pinterest’s spam detection system.
• Build Pinterest’s search function.
• Build Pinterest’s ad ranking algorithm.

Behavioral & culture fit

Unlike some other big tech companies—like Netflix, for example, which devotes multiple rounds throughout the interview process to evaluating an applicant’s culture fit—Pinterest mostly leaves the detailed behavioral interview questions for the on-site. (You’re bound to get a couple during the first recruiter call too, though, so don’t put off planning for them.) If you’ve come to the office for your interview, this “Values” round will often take place during lunch.

The canonical Pinterest values are:

• Put Pinners First. Product, business, and policy decisions at Pinterest are centered on the well-being of their users. Users’ voices are actively invited to the table when it comes to product direction.

• Aim for Extraordinary. Pinployees start with a high bar. They push themselves—and each other—to bring the courage, craft, and quality of execution needed to win big.
• Create Belonging. Divergent thinking, honest debate, and real-time feedback are understood as the fuel for innovation and growth. By reaching out to others, Pinterest members build strong connections and support the well-being of their people.
• Act as One. Pinployees put energy into helping their peers succeed, and their wins belong to the entire team. They tear down silos in how they work and rally behind the chosen direction after a decision is made.
• Win or Learn. Pinterest teams make big bets and take smart risks to increase the chances of impactful results. Whether they succeed or stumble, they ensure the learning is never lost by working hard to actively bring the lessons to our next effort.

Pinterest strives for a friendly and cooperative work atmosphere. Befitting their “Act as One” value, most project goals are assigned and evaluated for the whole team, rather than an individual. Your openness to collaboration, and your ability to receive and offer constructive feedback, are just as important as your technical expertise.

Sample questions include:

• Why do you want to work at Pinterest?
• Talk about a time you received critical feedback on a project. How did you respond to it, and what changes did you make as a result?
• Have you ever had a disagreement with a team member? How did you overcome it?
• Describe a time a member of your group didn’t pull their weight. How did you address it?

Additional resources

• Learn what it’s like to interview at Pinterest as a SWE, in their own words.
• Look over Pinterest’s engineering blog, open-source projects, and published research.
• Take our ML Engineering Interview courses to get a leg up on the topics and tools you’ll need to know.
• Book a mock interview with a Pinterest ML engineer so you’ll know what to expect.

FAQs about the Pinterest MLE interview

How should I prepare for a Pinterest Machine Learning Engineer interview?

To prepare for an ML engineer interview at Pinterest, practice answering ML system design questions. Read the Pinterest engineering blog to learn more about their product design philosophy, and study the projects you’re applying to work on. Check out Pinterest’s own SWE interview guides. Practice greenfield coding if you’ve gotten rusty.

How much do Pinterest ML Engineers make?

The average total compensation across software engineering levels at Pinterest are:

• (Entry-level) Software Engineer I:$228k
• Software Engineer II: $354k
• Senior Software Engineer: $469k
• Principal Software Engineer: $714k

How long is the Pinterest Machine Learning Engineer interview process?

The average Pinterest ML Engineer interview process takes about four weeks from the first recruiter call to completing the on-site. Some candidates zip through all three stages in just two weeks, or schedule each stage a month apart.