以前的几个面试

http://www.mitbbs.com/article_t/DataSciences/18319.html

发信人: gomakeitreal (benben), 信区: DataSciences
标  题: 以前的几个面试
发信站: BBS 未名空间站 (Sun Oct  4 01:28:44 2015, 美东)

以前面过几家，就给大家加几个data point.

note: no offer for these.


1) 资本1
- 4 coding online questions (can pick python or R or a few other languages),
nothing particularly hard. so don't remember anymore.
- phone technical interview (3-4 simple short questions, some are about your
projects and how you built your model; the last question do a groupby+sum
using 1 MR to check understanding of MR)
- onsite technical/case interviews (a lot of them are mentioned on mitbbs
already. pretty much the same. nothing new.)
- told fit is not good

2) 空气bnb
- take home exam: given data set, evaluate an AB test.
- onsite exam: 9-930 breakfast, 930-130pm work with 15 min lunch, then
present to a panel of 5 data scientist / pm. data very dirty. a lot of
cleaning and feature creations time. they emphasize on business insight,
which may or may not need sophisticated models. you bring your own laptop.
the exam is only ONE question: how to improve X based on the data. i heard
they have a set of questions to rotate, in case there is leakage.
- no reason given

3) 事件brite
- take-home + onsite: nothing really hard that i remembered, mostly past
projects, and since they wanted to build a prod reco sys. they asked that.
- no lunch provided. (this really sucks, b/c no one would have assumed that
if interview schedule includes noon time)
- they mentioned they are going to be more conscious and tries to look at
profit now. 聊了一下内部，不看好他们。 and not very sophisticated at all (in
terms of technology and models)
- recruiter not very friendly
- told fit is not good

4) 图钉
- take home exam: give you data set, do evaluation, like this is the data i
month 1 & 2, given data, anything really change?.
- did not pursue further