记录我这卑微的半年--经济学转ba应届找da/ds全职工作，内含21家面经

Job Hunting Interview
, , , , , , , ,
1

我是UT austin msba 强推这个项目
以下内容全部都是entry level的data analyst, 期待值高于这个的您将浪费您生命中宝贵的10分钟
楼主本硕都是国内经济学
硕士后公务员做了三年后出国（不相关经验），读了十个月的硕士项目
三月开始找工作，五月底（毕业了）开始全职找data analyst/data scientist工作，以上是背景
先列一下我面过的公司，时间顺序

  1. Facebook：内推onsite跪
  2. Amne ：海投onsite跪
  3. PayPal：内推 技术电面跪
  4. Indeed:两次hr screen后对方withdraw，后来知道indeed今年headcount很少很少
  5. ihs market: 海投hm面跪
  6. Neustar: 内推 二面（我也不知道算什么面）跪
  7. Via on demand: 海投 技术电面跪
  8. Shape security: 内推 hr screen后对方withdraw
  9. CUNA Mutual: 海投 onsite结果未出
  10. Offerup:内推 onsite跪
  11. Dimensional funds:海投 hr screen跪
  12. Poshmark:海投 offer
  13. Ask application:内推 hr screen跪
  14. Squaretrade:内推 回家作业跪
  15. Oscar health:内推 二面（debate question）跪
  16. AppDynamics:内推 offer
  17. Vixxo:海投，withdraw. check 1point3acres for more.
  18. Illumine:海投，录完视频面后跪
  19. AKQA：海投 offer
  20. Aicure:海投，withdraw
  21. Mckesson: 内推，withdraw

除了两三家大公司之外，全部都是小公司哈哈哈哈哈
因为我总计海投了500+，找内推找了小100
楼主心态很卑微，因为没有相关经验，项目也没有暑假所以不能实习，只求找到工作，要求很低
下面是面经加吐槽：

  1. fb全部都是论坛里看过的 不讲了，要点就是sql会challenge效率问题，applied data那一轮要详细到如何implement，所以lz讲的天花乱坠但不会implememnt,应该是跪在这趴了
    他家面试体验非常好，快！没有behavioral问题，没有尬聊过去经验，整个过程就像考试一样，准备的过程里lz把da的skill set都准备了一边所以跪了也不亏
  2. Amne: online test论坛里有，电面第一轮的基本的ml知识，处理outlier, missing data, 二轮讲相关经验，onsite是给数据集花一天时间写model预测房价，然后present,对lz来讲太硬盒了
  3. Paypal:印度大叔面我sql和挖简历，他讲话我只能听懂50%
  4. ihs markit:这家我想疯狂吐槽的，
    hiring manager问我理想的工作是怎样—我说喜欢分析data+吹对方公司的company mission;
    （他不满意），说先forget about this role, what roles have you applied for?
    ----（这就是丧命题吧？楼主只好老老实实说）fb, uber之类的
    这两个你更喜欢哪个？为什么更喜欢？
    ----（小孩子才做选择，我全都要！）我觉得差不多
    你为啥喜欢这两家，和别家有什么区别吗？
    -----（我申请了几百家，每一家我都要喜欢吗）。。。。
    总之聊的很尴尬，果然就跪了，这些recruiter真是很没有13数
  5. Neustar:这家也是奇葩，hr screen之后就失联了两周，我卑微又锲而不舍的催了两个星期，对方终于说在给你准备电面内容呢，约到了下周电面，也不告诉你面什么，经历了临时取消之后快一个月过去了，总算面上了，对方说很快，只要十分钟，结果第
    一个问题居然是问他家这套破系统的竞争对手有哪些，说越多越好 LZ傻眼了，，，不知道，对方问候也没有就挂掉了，直觉就是对方为了摆脱我终于想出了办法。。。。
  6. Via on demand:技术电面问的是纽约有一个地区乘客给的rating很低，怎么找原因,对方提示我怎么确定这rating是显著的低呢，lz于是想起来先做双sample t-test
  7. Shape security: hr忘记了我的面试，我发邮件给他才面上，之后hr消失了两个星期，邮件电话全失联，lz锲而不舍的去linkedin骚扰他，然后对方居然通过了，还已读了我的消息，不回，后来lz让内推人帮我去问问，之后终于等到了系统拒信
  8. Cuna mutual:这家hiring manager问的和hr题目一毛一样，，，而且感觉没有睡醒，也对我不感兴趣，后来讲好的onsite变成了视频‘onsite‘,可以 但没必要，真的没必要，继续是很水很水的面试，已经过去十来天了，还没有消息
  9. Offerup:技术电面问的sql是table for dealer & transaction, 想知道每个dealer每个月卖了多少，但是这个dealer不一定每个月都卖出过，这种情况下结果表要出现这个dealer这个月为0，lz的解决办法是join一个全部都是月份的table保证每个dealer都有记录，然后半小时后告诉我onsite
    Onsite是五个小时BQ，1两道sql,少量相关经验，五个人面我每人一小时，lz这辈子都没连续说过这么久的英语，bq主要有：
  • Experience driving toward a goal(challenges and overcome)
  • Experience you not meet the goal(define failure differently)
  • Previous projects
  • Experience when you use customer feedback to improve the status quo
  • Experience you strongly disagree with team(persuade others)
  • experience being scrappy, solve complex problem with simple solution?
  • experience when project fell off track (define failure, )
  • when you are primary project owner, what is the challenge(challenge and overcome)
  • experience building trust with teammate()
  • experience not fulfill the commitment with teammate()
  • experience pursuing excellence when you are unsatisfied with status quo()
  • Tell me about a time where you helped make your team stronger?(same with
  • How did you managed communicating with different group within the organization to implement a new change or policy.
    每个题都被问了两遍不止，地狱般的体验，再不想经历第二回了
    两道sql都是window function, 最好要熟悉snowflake的window function
  1. dimensional funds: hr问我有没有设计过使整个系统更有效率的办法？没答上来
    poshmark：电面sql+case

– table 1: tb_users
– 1. user_id
– 2. joined_at YYYY-MM-DD HH:mm:ss
– 3. gender
– 4. fav_brand

– table 2: tb_orders
– 1. order_id
– 2. item_id
– 3. order_at
– 4. buyer_id
– 5. seller_id
– 6. order_amount

– table 3: tb_items
– 1. item_id
– 2. item_brand
– 3. item_gender

/* I want to know if the second item sold by a user is also their favorite brand. I want this for all sellers */
select t1.id, case when t1.fav_b=i.item_brand then 'yes' else 'no' end as '2nd_item_fav_brand'
from(
select u.user_id as id,u.fav_brand as fav_b, o.item_id as iid , rank() over (partition by u.user_id order by order_at) as rk
from tb_users u left join tb_orders o
on (u.user_id=o.seller_id) ) t1 join tb_items i on (t1.iid=i.item_brand)
where t1.rk=2;

/* Can you also add amount they sold in 2018 to the query below */
/*Im a marketing manager. I want a list of all our users: user id, date they joined and amount they purchased in 2018*/
select u.user_id, date(u.joined_at),sum(case when year(o.order_at)=2018 then order_amount else 0 end) as 2018_amt
sum(case when year(o1.order_at)=2018 then o1.order_amount else 0 end) as sell_amt
from tb_users u left join tb_orders o
on (u.user_id=o.buyer_id)
left join tb_orders o1 on (u.user_id=o1.seller_id)
group by u.user_id,date(u.joined_at)

case是total sales/orders这三个月下降了

  1. external: seasonality?(after Christmas or other holiday )
    all of a sudden(tech glitch)
  2. from formula: how total sales change(customer of value churn)
    how total orders change(cheap order increase)
    distribution of order value to dive deep in 2
  3. from funnel: adoption/engagement/task success rate/retention/user experience
  4. from segment: new and old/male and female/platform/catogory
  5. from competitor

终面还是视频面 四轮，实在想不起来了，又一轮特别freestyle,就记得整段垮掉了