I've recently been interviewing for a Senior DS role at Discord. I just passed the hiring manager and technical phone interviews and am preparing for the virtual onsite.
HM:
Introduce yourself.
Behavioral question: Tell me about a time you used causal inference to solve a business issue. What was the challenge, and what would you do if you did this project again?
At the end there was a small case: The company is trying to launch a new version of a feature based on a classification model. Other than an A/B test, what other methods can you think of to measure the success of this model?
I interviewed in the morning and found out I passed that afternoon.
Technical phone: case + Python, 1 hour.
Python:
Implement number_of_character(string, char) to count how many times character char appears in string. Case-sensitive.
Implement unique_string(msg1, msg2). Given two message strings msg1 and msg2, return a sequence of words that appear in exactly one of the two messages, the symmetric difference. Output a list.
I had to run both and show the results during the interview.
Case: We have a model that identifies which users might subscribe, and we need to send notifications to users who are likely to sign up for the subscription service.
Measure success: design a metrics framework, etc.
Design the experiment.
Follow-up: Is there anything to do before the experiment? An A/A test.
If the result comes back flat, what should we do?
Are there ways to experiment other than A/B tests? What are the trade-offs? List at least two or three options with pros and cons.
Follow-up: Deep dive into one of the options.
Job hunting is hard!
Discussion
Loading comments…