I applied through an employee referral. I interviewed at Uber (San Francisco, CA) in Jul 2017
Interview
Referred through a friend, had a few quick informational calls then was sent the take home assignment, followed by onsite.
The take home was a decent problem though it took a while, and I was told I "nailed it". They made me create some slides about it which I thought was strange because the take home assignment was in presentable format, and when I got there it seemed like no one had really looked at my take home. People were nice to me, but scheduling was a mess; they kept on pushing back and moved my onsite at the last minute.
I didn't get asked that many technical questions when onsite, but I got feedback saying my ML/modelling was a concern. Given I had to do a lengthy take home modelling exercise, I am confused but I suppose there could just have been more experienced candidates.
I applied online. I interviewed at Uber (Toronto, ON) in Jun 2017
Interview
Apply the data scientist job online. First a HR will call you to arrange a phone review. In this phone review will ask you some personal information and your background, then the HR will help you arrange the interview with a manager. After the phone interview with the manager, they will let you know if you are hired.
I applied through a recruiter. I interviewed at Uber (San Francisco, CA) in Oct 2016
Interview
preliminary phone screen with a recruiter, technical phone screen, take home assignment, and onsite interview which had six segments: hiring manager (focus on culture fit and soft skills), two on quantitative data science (focus on applied statistics, modeling, and experimentation), whiteboard coding (in Python), product manager (focus on communication skills and products), and someone outside the data team (focus on company fit).
Interview questions [1]
Question 1
- talk about a recent project that was particularly challenging and why
- metrics to evaluation surge pricing algorithm
- how to test whether version 1 or 2 of surge pricing algorithm is working better (divide drivers - how would you explain power to a product manager (non-stats)
- basic modeling questions: cross-validation, training / test data, error metrics