Pinterest Data Scientist Interview Questions
& Process

I had an online assessment with 4 questions, which included 2 SQL queries and 2 Python questions focused on data cleaning. I managed to complete 3 out of the 4 questions, passing all the tests, but it still wasn't enough to move forward. If you pass this stage, there are typically 3 more technical rounds.

I had an initial chat with the hiring manager, followed by a coding screen, and then an onsite interview. I thought I did well, but they said I didn't answer one question in depth, even though I offered multiple solutions. The recruiter mentioned I 'did a poor job going in depth' and that offering two solutions was seen negatively as 'switching up.' It's interesting that a recruiter, who wasn't even in the interview, would give such feedback, but it reflects the subjective and arrogant culture at Pinterest. It felt very subjective, and I suspect it might be a company with DEI hires for recruiters who then act abrasively if your interview doesn't directly lead to an offer for their benefit. The interviewer who gave this feedback was apparently Caucasian and might have been looking for a specific type of candidate, leading to a potential bias in their evaluation. It seems like an unwelcoming environment unless you fit a very specific, narrow mold – characterized by racism, arrogance, and entitlement. I'd recommend avoiding it.

When I applied for this data scientist job, the interview process kicked off with a quick recruiter screen where I walked through my background and tried to sound naturally enthusiastic while also praying they wouldn’t ask me to define MLOps in a sentence. Then I moved on to a technical phone round, which turned into an unexpectedly deep dive into pandas operations and left me second-guessing everything I thought I knew about DataFrames. After that came the “simple” take-home assignment that somehow ate my entire weekend and required me to dust off parts of linear algebra I hadn’t touched in years. Finally, the onsite loop involved a whiteboard session where I derived logistic regression under the mildly judgmental stare of three engineers, a product case where I had to improvise strong opinions about churn, and a behavioral round where I repeated “cross-functional collaboration” so many times that I briefly wondered if I was still speaking English.