I applied online. I interviewed at LinkedIn (Sunnyvale, CA) in Dec 2020
Interview
Nice format and interviewed by a gently person. Some questions about product and basic SQL. Beware of the syntax and formulate your answers. It is great that I receive a feedback from the HR manager.
Interview questions [1]
Question 1
How would you write a query to obtain "X" result? How would you find people on Linkedin that works in "X" industry?
I applied in-person. The process took 5 days. I interviewed at LinkedIn (New York, NY) in Sep 2019
Interview
You need to understand that a data scientist role in companies like LinkedIn or Facebook has nothing to do with data science in a way you imagine it. You will not be building models, you won't be training models, you won't be updating models. Most likely, you won't be even doing A/B tests. All you need to know is SQL and how to do self joins. In the industry, once a company sees that you're a data scientist at LI or FB (not a researcher), it immediately understands that you don't possess any required skills to be a "real" data scientist. If you want to be a data monkey with a good pay but without a single chance to move to a real DS role, go for it.
Interview questions [1]
Question 1
Walk through the resume; what is one feature you want to have added; how will you assess the success of this feature
Tech part: exactly the same SQL questions like in Facebook's interviews (left join, self join) but then was asked to rewrite everything in Python
I applied through an employee referral. I interviewed at LinkedIn
Interview
Applied through a referral. Shortly after, a recruiter responded to set up the first round of the interview process, which was a technical screen.
This was done in SQL and Python through a shared text editor (that is, both the interviewer and interviewee can see what each person is typing). The questions were quite easy and very straight forward. Made a few minor slip ups here and there (something that can easily be caught in a real-life situation when you'd be able to see outputs or error messages). Here's where I believe the assessment was unfair. There was no specification on which variant of SQL this interview would be conducted in. I coded up my SQL commands in PostgreSQL, but there are slight variations in some of the functions/syntax across different SQL languages. Despite that what I coded was 100% correct in PostgreSQL, because it didn't work with Presto, the interviewer deemed it incorrect.
The interviewer was also clearly copying the query I wrote into the editor to verify, so that's why she must have thought I was wrong. So despite writing a completely working query for each of the questions, because it did not conform with one specific variant of SQL, I did not make it to the next round. The number of changes needed to even get it to a working state were so minor too. At least 95% of the query was exactly as she expected it to be.
Honestly I was very surprised I didn't move forward... It seems she expected everything to run 100% correctly with no typos or errors despite me not being able to run the code to see the output. Absolutely unfair interview process.
Interview questions [1]
Question 1
Basic SQL manipulation with joins and group bys.
Python pandas manipulation