lyft data science onsite interview overview for software engineer position

First Round, Statistics

先写了 sql / pandas 几个问题

然后问了下面的统计问题

Context: After successfully launching targeted couponing, the team now wants to test using coupons to encourage riders to use Lyft for commuting. Suppose there are N commuters who take two trips per day, one trip to commute to work and another to commute back home. The probability they choose Lyft for each trip is:

• Probability of choosing Lyft to work = p, where p < ½  

• Probability of choosing Lyft going home = 2p IF they rode Ly...

First Round, Statistics

先写了 sql / pandas 几个问题

然后问了下面的统计问题

Context: After successfully launching targeted couponing, the team now wants to test using coupons to encourage riders to use Lyft for commuting. Suppose there are N commuters who take two trips per day, one trip to commute to work and another to commute back home. The probability they choose Lyft for each trip is:

• Probability of choosing Lyft to work = p, where p < ½  

• Probability of choosing Lyft going home = 2p IF they rode Lyft to work, otherwise it’s p

Question 1: What is the probability of a rider commuting back home with Lyft?

Question 2: If a rider takes Lyft home, what is the probability they took Lyft to work?

Final Rounds

第一轮 ML system design, passenger churn prediction，地里有原题，就是给一个data schema，聊一下怎么 build model 预测 user 在接下来几周会不会 churn

第二轮 coding, implement stratified k-fold cv

第三轮 business case study, 说如果Prime Time在1-3周的weekend都是stable的但是第四周突然变大，该怎么分析root cause。

盲猜第三轮不行，HR反馈不面第四轮了