Pinterest Data Scientist Interview Guide

This interview guide has been written for people preparing for Pinterest’s data science hiring process. We break down what each round covers and what Pinterest interviewers are looking for. Our writing has been informed and reviewed by senior data scientists working at Pinterest.

Pinterest calls itself a visual discovery engine. The main task for data scientists at Pinterest is figuring out if a Pinner found something that inspired them, and sending them the next Pin that keeps that spark alive.

A Pinner could be looking for a hiking backpack, find inspiration Pins for a hike in Machu Picchu, then see images of Peruvian food, and end up planning a trip to Peru.

As Tamara Louie, a Pinterest data scientist described on the Pinterest Engineering blog, Pinners can travel hundreds of levels on the ‘Pinterest Path’ deep in a single session.

TL;DR Pinterest’s data science interview loop has 3 parts:

Step 1: Recruiter screen

Step 2: Technical screen

Step 3: Onsite

• 3.1 Coding/Python
• 3.2 Statistics and probability
• 3.3 Product and metrics case
• 3.4 Specialization 
• 3.5 Behavioral

Pinterest's DS interview has three phases: a recruiter screen, a technical screen with a hiring manager or Pinterest data scientist, and a four to six round onsite loop. 

The interview process kicks off with a standard recruiter screen. Their job is to decide if a candidate’s background and skillset fits the open positions and if they clear the bar for a technical screening. 

Recruiters will talk in detail about the roles and responsibilities of the position along with Pinterest’s culture.

What questions should you expect? 

• Can you explain your background?
• Why do you want to work at Pinterest?

You will have an hour-long technical screen either with the hiring manager or a data scientist from Pinterest.

Overall there are 3 parts to it:

1. The first 15 mins will focus on your past work and projects.
2. Then SQL & Python (or even R) questions.
3. Finally interviewers will ask statistical and experimentation questions.

For the coding part, expect a bigger focus on SQL, because there’s no SQL round in the onsite nowadays.

SQL questions for Pinterest are based on their users' sessions. Usually the datasets are super messy and will require you to go through a few clean up steps before you can start analyzing.

Pinterest DS candidates have been asked these questions:

• Assume you are given the below table on sessions from users, with a given start and end time. A session is concurrent with another session if they overlap in the start and end times. Write a query to output the session that is concurrent with the largest number of other sessions.
• Write a SQL query to count the number of unique users per day who logged in from both iPhone and web, where iPhone logs and web logs are in distinct relations.
• If you have a database table with column user_id, date, event_id, so event_id will be a user saving a Pin. How will you get the streak count for all users at all dates? A streak is the number of consecutive days a person has saved a Pin.
• How do you find the right sample size and choose metrics for an experiment?

The onsite loop will have four to six rounds in total. Most interviews will take 45 mins, the exception is the case round which will be an hour long. 

Product knowledge and sense is super important for Pinterest, and you might be asked Pinterest-specific questions in any round. This Reddit commenter noted Pinterest interviewers ask questions related to their own architecture and Pin board system. 

You will be tested on your Python and especially pandas abilities. Expect to solve two problems in 45 mins.

The most common type of problems candidates report relate to string manipulation and array operations.

If you’re interviewing for an ML-specific role or on their Trust & Safety team, expect more difficult questions.

Pinterest DS candidates have reported these questions:

• If you’re given an unsorted array of decimals, how would you find the interquartile distance?
• How would you write a function to find the maximal shared substring of two strings?
• Show me how you’d implement a SparseMatrix class with construction and basic operations.
• Can you do weighted random sampling without using numpy.random.choice directly?
• Can you create a function to get the cumulative sum of a list?

If you see a more optimal approach after you've written a brute-force solution, say so. Pinterest interviewers want to see how you think as much as they care if you’ve got a working solution.

You will be tested on experiment design in two separate rounds (this one and the following product case round). 

First, the stats round will focus on how strong your knowledge is. Expect more theoretical and less practical questions here. You could also be asked about something related to your projects.

Some statistics and probability questions candidates have been asked recently include:

• Describe an A/B test from start to finish.
• What's the mathematical intuition behind logistic regression?
• Can you derive logistic regression?
• Describe and implement bootstrap sampling.

For the product round you’ll be given a Pinterest case and will combine your experimentation knowledge with your product intuition to walk interviewers through the problem end-to-end.

Your interviewers will walk you through a case deck and will also expect you to come up with metrics.

In rare cases, this round could also feature a question (or followups based on main prompt) typically asked to data engineers.

Pinterest interviewers want to see how you apply statistical knowledge to make calls even if the data isn’t clear. They’re also evaluating how you communicate your reasoning for different levels of technical knowledge your colleagues might have.

This round will be more conversational and there’s likely no right answer to the questions. You’ll be expected to define and break down the problem in detail before going ahead.

Product and metric questions reported by Pinterest candidates:

• How can we measure if the feed shown to the user is diverse enough?
• We want to add a horizontal carousel at the top of the feed. How would you design the A/B to test if it’s a good idea? What are the metrics you’d choose?
• How would you validate brand lift on Pinterest ads?
• Can you explain what you understand about Pinterest’s ad recommendation system?
• How do you manage the complexity of a pipeline with a lot of transformations?

A Prepfully coach who is a senior data scientist at Pinterest says:

Your approach matters, but interviewers are evaluating how deeply you understand Pinterest and its user base. 

If you’re defining success, on top of engagement metrics also talk about the ‘Pinterest Path’ we wrote about up-top. Are we showing Pinners something new and cool they didn’t even know they were looking for?

We highly recommend you spend a lot of time using the app before your onsite. You should know the basic terminology and also think about how you could improve Pinterest for any specific interests you might have. Your first-hand personal opinions will make answers stand out to interviewers.

Before your onsite, Pinterest asks you to choose a specialization in one of these three: Statistics, Machine Learning, or Forecasting. 

You might have one round that will be tailored to that. It could be its own separate round or be a part of the product case round we’ve described above.

Candidates usually get three questions, starting with a high-level scenario that’s specific to your specialty. Pinterest interviewers want to see how you work on ambiguous problems. 

Pinterest interviewers can ask DS candidates these types of questions:

• Let’s talk about the most complex ML problem you worked on.
• How do you choose when to apply L1 or L2? How does regularization work for both?
• A fair die is rolled n times. What is the probability that the largest number rolled is r, for each r in 1..6?
• You are given an unfair coin, with an unknown bias towards heads or tails. How can you generate fair odds using this coin? (the classic Von Neumann trick)

If you pick stats, expect a big emphasis on experiment design and causal inference methods. If you go with ML, expect applied ML problems. And if you choose forecasting, you’ll have to solve time-series problems featuring Pinterest-specific metrics.

To understand the depth of your technical experience, interviewers will also ask you about your past work. You’ll be expected to provide the process and reasons behind any decisions you made.

We recommend you jot down your thoughts and calculations as you go on the whiteboard or a notes app. It’ll help you remember specifics and give your interviewers a chance to redirect you if they want.

Interviewers are asking: “Does this candidate fit into our culture and will they add value to the team and organization?” And really, they just want to know if they’d enjoy working with you.

You should expect a lot of questions focused on collaboration and cross-functional work. Most questions will be about your past work but you could be asked a few hypotheticals as well.

Pinterest interviewers have asked these questions:

• Can you tell me about a time you had to explain a super complex technical detail or analysis to a non-technical colleague or stakeholder?
• Tell me about a time when you used data to influence or persuade a decision that another person had to make.
• Tell me about a project where you got back surprising results. Why did that happen?
• How do you make sure your team is in sync with any other teams you might be working with?
• Have you ever relied on your intuition for your work?

While ‘the why’ is important, interviewers also want to hear details about the interpersonal side and human relationships you had to manage. So how did you talk with this stakeholder, what did you do when your request was denied, and so on.

Total comp for DS roles at Pinterest ranges from $205K at L3 to $462K at L6, according to Levels.fyi.  The base range is from $157K to $238K.

Pinterest's RSUs vest either on a 3-year schedule (50% in year 1, 33% in year 2, 17% in year 3) or a standard 4-year 25% vest. According to Pinterest candidates, there’s room for negotiation, so don’t shy away from it.