【在线Info Session】业界数据科学家回答你关于找数科工作的问题

楼主 问一下 ms想要找ds的工作一般需要掌握哪些fundamental knowledge呢 需要会数据结构 刷题再学习ml和dl嘛

小K 发表于 2017-5-28 07:21
I am a data scientist and not an analyst.
DA: 用不到任何算法，一般要求是SQL和Excel吧，如果已经到 ...

只会Python有钱途么？我马上要读个DS的MS，偏统计，data engineer需要什么样的技能？spark什么的么？我只会Python，需要再去学点什么么？
女朋友读的analytics，我看她的课程很多data什么的，她比较想做广告推荐之类，这个专业出来能做么？推荐这东西大概需要多大的知识才能hold住？
还有DS对于学位有什么要求么？analytics出来能做DS的工作么，如果自己够努力的话？

arieslpx 发表于 2017-5-27 15:54
楼主 问一下 ms想要找ds的工作一般需要掌握哪些fundamental knowledge呢 需要会数据结构 刷题再学习ml和dl ...

之前只有讲过108遍这个问题而已。谢谢

BIUBIUBIUBIU 发表于 2017-5-27 16:23
只会Python有钱途么？我马上要读个DS的MS，偏统计，data engineer需要什么样的技能？spark什么的么？我只 ...

牛人永远有钱途。。。。这还能说啥呢。。。。

analytics做推荐系统没戏

在申请之前，看了K姐数据科学家炼成的那个帖子，一直也在跟着学，后来去投简历发现analyst bar 还不高， ds就很高了
于是转码了。。。

PhD还看学校是什么么？ds一般都是名校PdD 还是都有？

1. 楼主说的ship fast路线的有哪些公司或者职位可以介绍下吗？
2. Master长期想做比较技术的DS，大tech公司的DA和小公司但是能做DS（不是挂个名，可以用package做模型应用，技术接触面较广），楼主认为哪个作为起步更好？有什么要考虑？

感谢！

楼上/上这些问题的答案只能是 it depends....
Yes a top 5 rank definitely helps you in subtle ways --- but that's not how ppl decide things, why would anyone choose to go to a lesser ranked school if everything else is equal? they wont, because everything else is never equal.

nothing is suitable for everyone, without proper context ,it is impossible to say which is strictly better than which else.

siciliana 发表于 2017-5-28 16:48
1. 楼主说的ship fast路线的有哪些公司或者职位可以介绍下吗？
2. Master长期想做比较技术的DS，大tech公司 ...

2 小公司建模更好。
小公司的确接触面更广，作为新人可以更快成长。但是前提是需要好的mentor，如果全公司就一个DS，可能做错了都没人能告诉你

缺点可能就是，新人需要人带。
个人认为DS是个apprenticeship model，一个牛DS也需要mentor来指路，最好还需要有a team of good DS to bounce ideas with. 如果去好的组，1-2年内可以学到下面5-10年用的内容和思路。


考虑小公司的时候，建议考虑如下问题：
他们需要DS吗？还是只需要DA? or are they thinking about data eng? 他们知道这些什么区别吗？真的知道吗？
他们公司的stage真的需要DS吗？ 真需要建模吗？还是随便slice and dice就够？
谁负责派活儿？那人懂DS吗？他什么出身？之前跟DS和DA和eng合作方式如何？
你建模打算如何demonstrate business impact? 你认为的方式他们认同吗？
他们的数据pipeline成熟到能支持建模的地步吗？DS这种东西太容易是garbage in garbage out. Without a mature data pipeline, data science is just a sham. Who are their data engineers? how long has their data eng team been in the field of data engineering? What is the scale of things they have worked with? do they understand typical problems that can happen in a data pipeline? do they understand data leakage? Other than "git pull" airflow, what else have they tried? what problems have they seen in the past 12 months? how did they fix it? what is the rate of data loss? how did they check it? Did they even check? How are schema's enforced? What tech stacks are they using? how are these chosen? Given their data scale, does that make sense? What else have they achieved in the past regarding data eng? 问这些，是因为太多的小公司data部门就是几个小本随便搞搞，各种break各种loss，这种情况下做个P的science.

他们PM会随时找你问问题吗? are you expected to perform ad hoc queries for them? Do their PMs know SQL?

Are DS expected to participate in daily scrum? Really, daily??
Who decides what projects to take on and when to deliver them?
Is there anyone with a PhD (in any field) in the company? (Hard to do science/modeling work, if there is not a single PhD, in any field anywhere)
How much production code do they expect their DS to write? Do DS participate in pagerduty? On call?
What stack do they use for DS work? If the answer is Java, dig really deep on what kind of modeling they are doing.