Table of Contents
Pinterest, Inc. is a social media web and mobile application company founded in 2009, headquartered in San Francisco, California. The company develops and operates software applications and systems, designed to enable the discovery and saving of information online using images, GIFs, and videos (known as Pins). It offers free registration, after which users are allowed to upload, save, sort, and manage images and other content, eg videos (pins), through a gallery of images known as pinboards.
Its average user-ship has grown steadily since its inception, as audiences frequently turn to the platform for “planning social activities, shopping, learning things through how-to posts, and planning out life’s moments with boards for visual inspiration”.
As of the 4th quarter of 2019, Pinterest's active average monthly users crossed 335 million worldwide, with over 175 billion items pinned on over 3 billion virtual pinboards. With this information, it is not far-fetched to imagine the massive amount of data generated daily. Data Science is at the core of Pinterest products and services, and data scientists at Pinterest leverage the most advanced analytics tools and machine learning models to make sense of this data for guiding business decisions.
The Data Science Role at Pinterest
Even now, Pinterest is still a growing company with many teams and departments working on key features, products, and services for improving customer experiences.
The data science team at Pinterest occasionally collaborates with other teams to design experiments around almost every user-facing feature to help make sense of the huge customer data generated daily, driving decision making and providing business-impact insights. As a result of this, data scientist roles at Pinterest are hugely determined by the assigned team. However, general data scientist roles at Pinterest span across experimentation and statistical modelling, basic business analytics and data visualization, machine learning and deep learning theories.
Interested in data science at another comapny with huge amounts of user data? Check out "The Deloitte Data Scientist Interview" article!
Pinterest hires only qualified Data Scientists with at least 3 years (6+ years for a lead role) of industry experience in relevant data science projects. Requirements for hire are very specific depending on the job role for the team and as such, it helps to have specific industry experience that aligns with the role on the team.
Other relevant qualifications include:
- Advanced Degree (MS or PhD) in a quantitative field or related fields.
- 3+ years experience (6+ years for a senior role) of industry experience and a proven track record of applying statistical methods to solve real-world problems using big data.
- Industry experience in both online and offline experimentation.
- Experience managing and analyzing structured and unstructured data with SQL, R or Python, and using software packages like SPSS, STATA, etc.
- Extensive experience with applying deep learning methods in settings like recommender systems, time-series, user modelling, image recognition, graph representation learning, and natural language processing.
- Experience with learning from ranking labels (i.e. triplet learning, metric learning, etc.) and deploying ranking models (i.e. learning-to-rank).
- Ability to lead initiatives across multiple product areas and communicate findings with leadership and product teams.
What are the data science teams at Pinterest?
Data scientist roles and functions at Pinterest run across a wide range of teams and fields related to data science. The title “data scientist” at Pinterest encompasses multiple roles and functions ranging from product focused-analytics to more technical machine learning and deep learning functions.
Based on the assigned team, the function of a data Scientist at Pinterest may include:
- Engineering (Offline Experimentation): Leveraging advanced data analytic concepts to solve key measurement challenges involving the offline evaluation of data, from fine-tuning measurement techniques to defining approaches for creating meaningful measurements of value for new and existing new products.
- Engineering (Ads Experimentation): Designing and building models, mechanisms, and metrics to make sound product decisions through experimentation with the end goal of surfacing high-quality ads for every Pinner.
- Business Operation and Strategy: Leveraging business analytics to drive critical business insights for a better understanding of Pinners, Partners, and products.
- Ads Quality Ranking team: Applying experimentation, quantitative analysis, data mining and data visualization techniques to improve the quality and relevance of ads on Pinterest.
- Ads Intelligence: Developing machine learning models, systems, and features that help advertisers maximize the return on investment of ad campaigns on Pinterest through recommendations, tools, and insights.
The Interview Process
The interview process starts with an initial phone screen with a recruiter or a hiring manager, and if all goes well, a technical screen with a data scientist or a data engineer will be scheduled. After passing the technical screen, you then proceed to the onsite interview, which comprises five back to back interview rounds with a lunch break in between.
This is a 30 minute initial phone conversation with a recruiter, detailing your technical background, your past relevant projects, and a quick assessment of your skill sets based on your resume. Within this interview, the interviewer will also discuss with you the roles on the team and Pinterest culture.
- Tell me about yourself.
- Talk about one of your past work experiences.
The technical screen is an hour-long interview with a data scientist, with discussion revolving around a past project, the approaches you used, and how you solved certain challenges.
There will also be some light SQL coding in this interview. Pinterest uses “Karat” for almost all their technical interviews and the Data Scientist technical screening is also done using the shared screen Karat platform.
At a minimum we recommend reviewing this article about "Three SQL Concepts you Must Know to Pass the Data Science Interview" on Interview Query to prepare for your interview.
The onsite interview is the last interview stage for the Pinterest Data Scientist interview. It consists of five back-to-back interview rounds, split between a SQL interview, a statistics and probability interview, one coding interview, and a behavioral interview. All interview rounds in the onsite stage last approximately 45 minutes, with a lunch break in between.
Notes and Tips
Pinterest Data Scientist interviews aim to assess candidates’ ability to design experiments for assessing product performance, build models at scale, and apply data science concepts to drive growth and provide business-impacts insights. Therefore, interview questions are standardized and cover a wide range of data science concepts. Brush up on your knowledge of statistics and probability, hypothesis testing, time series modelling, A/B testing, experimental designs, SQL, and predictive modelling concepts.
Practicing interview questions from Interview Query can better prepare you for the technical aspect.
Pinterest has an employee-focused ecosystem, which provides a friendly work environment for all. In a 2019 article, Pinterest was quoted as “the nicest company in Silicon Valley… The culture stands out from other high-growth tech companies where confrontation and debate are actively encouraged”. Culture-wise, Pinterest offers a really progressive work environment where employees (technical or not) can grow and thrive.
Another company with great work culture is LinkedIn. Check out this guide about "LinkedIn Data Science Interview Questions".
Pinterest Data Science Interview Questions:
- Give an array of unsorted random numbers (decimals), find the interquartile distance.
- 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.
- Your product manager noticed a dip in a specific metric. How do you go about investigating what may have caused the dip?