黑车DS店面+昂赛挂经

店面一:
Why the company? Tell me about your favorite ML project? What is the shortage of your method? How to improve it ?
Case study: How to predict churn behavior on the driver side?
python OOP question: searching order of inheritance in python?

电面二:
SQL: easy, follow up: what’s the difference of where and having in sql?
ML: explain bias/variance trade-off, how does bias/variance change if you add penalty term in regression? How about reduce the depth of decision tree?
What is ridge regression and lasso regression? What is different use of these two?
Python or R:
Implement a simple linear ridge regression without package
How to adaptively change learning rate? Implementation

Onsite:
1 Bar Raiser: Behavior question + Design a surge price optimization model? + How to evaluation if a surge price method is good?

2 PM: A/B testing, 问的特别细

3 coding, 1道leetcode简单题; 一道data manipulation的,可以用sql/r/python

4 很多product sense + case study。大概有how to evaluate if uber’s market is supply-constrained or demand-constrained? If you can only track on metric to evaluate if Uber’s growth is health, which metric would you choose? Why?

5 详细问你的ML project, 会follow up问这个model的basic,要求一部分白板推导

6 hiring manager, 全部是behavior question

请问楼主的面的是data scientist吗?
uebr 的data scientist不是要phd吗?

是DS。对的,大部分是PHD,但也不完全都是。楼主的面试官也有2个master。