IMO these fields can all be Keng if data is not properly used.
Of course, MLE can be better. after all fields like NLP is kind of more mature than other data sciences' subfields so you can see better understandings/results there. Since you have engineering background it may be easier for you to appreciate it.
I am actually a PhD in mechanical engineering. How likely it is to find a DS/MLE job with a PhD degree + 1 year industrial experience? Most ME PhD graduates who pursue a career in Mechanical Engineering only make ~100k year. I would be pretty happy if I am able to make more than that.
Well at least I can find a job in data science. Mechanical engineering job market is shrinking in the U.S. and China. Even my American classmates can't find a job.
I am actually a PhD in mechanical engineering. How likely it is to find a DS/MLE job with a PhD degree + 1 year industrial experience? Most ME PhD graduates who pursue a career in Mechanical Engineering only make ~100k year. I would be pretty happy if I am able to make more than that.
My team has 4 data scientists. None of us gets promising results so far. That's why I get frustrated. Have you worked with consumer data before? Could you recommend any good learning resources? Thanks!
我是传统工科转行data scientist,现在在一家IT consulting公司做了几个月左右,工作内容是consumer behavior predictive modeling。老实说我觉得这和生物纳米一样就是个坑。机器学习模型预测结果的准确率只比乱猜稍微高一点,但是公司靠玩弄一些数字游戏把结果变得很好看然后去客户那里骗钱。长久下去,客户迟早会意识到这个根本没用。不过我在工作里还是学到了一些比较时新的技术比如nlp, deep learning之类,所以我现在打算转行到data science比较靠谱的领域。
data science下面有几个方向: risk assessment, fraud detection, machine learning engineer等等,大家觉得哪个方向相对偏技术,职业发展好一点?我听说risk assessment好像也是坑。如果掌握一些编程技术的话,做machine learning engineer会比较好,因为可以做产品。大家怎么看?
Of course, MLE can be better. after all fields like NLP is kind of more mature than other data sciences' subfields so you can see better understandings/results there. Since you have engineering background it may be easier for you to appreciate it.
Thanks for the input! Sounds like only PhD in CS/EE/STATS can be qualified for these high level MLE positions.
I am actually a PhD in mechanical engineering. How likely it is to find a DS/MLE job with a PhD degree + 1 year industrial experience? Most ME PhD graduates who pursue a career in Mechanical Engineering only make ~100k year. I would be pretty happy if I am able to make more than that.
数据不好吧?不是随便把数据放model里跑跑就是data science了,consumer behavior model可以很复杂很有用。
Well at least I can find a job in data science. Mechanical engineering job market is shrinking in the U.S. and China. Even my American classmates can't find a job.
努力刷题是王道 ---发自Huaren 官方 iOS APP
MM可以说说自己转DS的经验吗?我和你背景类似,正在转。
我自己的思考是可以从两方面增值:
一是做公司的核心业务,比如在电商零售做推荐系统,在石油行业做故障预测,在今日头条这种公司做核心算法等
一是做全栈DS,从数据获取,分析,算法,到数据产品的deploy都懂得话,那自身肯定是大大增值了。
My team has 4 data scientists. None of us gets promising results so far. That's why I get frustrated. Have you worked with consumer data before? Could you recommend any good learning resources? Thanks!
You are making the same mistake as having studied mechanical engineering by preferring something that's more technical and "solid".
我LinkedIn上面有一个人跟你一样的专业,成功转做machine learning engineer 了,不过他有刷kaggle (排名比较前那种),在LinkedIn积极写东西(nlp computer vision 之类的博文)在LinkedIn有两万多个follower😂,有maintain比较好的github portfolio ,当然自己挺费时间的,我看他大概花了一年多时间才有那么多人关注
Re.
预测不符合预期,无非两个原因。model不对,或者数据不对。
真正的ds应该会根据实际情况选择至少tune model,采集合适的数据作predictor.知道为什么这么采集。然后还要会解释结果。达到这个水平一般要起至少有cs相关方向和统计相关方向的phd加多年的industry 经验。
如果就拿现成的软件,跑跑别人采集的现成的数据,不知道为什么。结果和预期不符合还不知道如何解释。根本不是真正的data scientist
不过好多职位也是乱写。好多能用R就叫data scientist了。实际上连入门都算不上。
不要phd啊。我有个同学cs master,本科物理,在狗家做nlp engineer。
写代码和做研究两码事。能看懂帮researcher实现出来也有这种岗位。
太扯了。
Exactly
不是技术好就是牛,是这条路比走business路线容易多了。偏business方向第一要有sharp vision,第二要有很强的communication skills,要有极强的story telling的能力,第三要有很好的inter personal skills,缺一不可
为啥这些大家不愿意学?5年PHd都拿到了,这点还不是三年就好,加起来9年也搞定了。
这类工作特别有impact有意思。建议找个数据多的公司干,金融和高科技都是不错的领域。
就是啊,怎么会没用,model经常要调整的,和行业也密切相关,不是软件出来什么就是什么。如果只比随机的好一点点,就要找问题在哪里
还真不是想学就能学出来的。这些softskill比那些技术上的skill set难提高多了。很多技术很强的人不愿意或者不能做management track,为什么?就是soft skill上不去,但是做technical work只会干活儿也是能混下去的。在business side,这些soft skill都是必须的,不然都没法survive。并且不是所有人都喜欢business side的,很多时候理工科思维也接受不了business side的思维方式。
做IC本身收入也不错,而且我觉得管人真的挺累的,要跟其他组撕逼,单纯就不喜欢,我就喜欢做IC,像我男朋友喜欢跟人开会聊天,不喜欢写code他就转走management路线了
一般有这几样能力的都混得很好,我前老板就是,技术好,会说,会写,懂business,这几样确实很难一下学会,可能还需要点天赋,或者要大步走出舒适区。
半路转行要么做点虚的偏business的东西,要么做点简单编程那种随便一个小本科都能做的
我做教育这一行,量化方向。现在也是流行DS,建模,有些同行是从完全非相关领域 转过来做DS,完全不懂教育,建的model可以说一点用处没有。
你数学咋样? 学过什么课?
评估一下
类似的困惑来看答案的
fraud 和 risk是我觉得饭碗比较稳的两个方向 但是各自有各自的局限
fraud 如果不是payment chargeback 有data sparcity的问题
risk 有compliance 问题 模型都需要有可解释性
nlp什么的实在是聪明人扎堆 可能要比其他方向更多的努力
透露下你们公司
看起来养老不错
我也赶快去申请
她soft skill好啊也好看,她说话说一大堆说的慢,别人都不打断。
她就是什么都做不出来,也能说一大堆,而且还表现的理所当然,特别自信,好像做的那东西有多难似的。
这一点,学不来啊
我们行业里钱最好的一个是alpha research,一个是marketing。
market risk是其次的。
我们组DS们也做,但是目的并不是predict,而是在这个过程中offer 2nd opinion,提供业界一些insights,最后服务于economic capital,直接影响股票价格,所以CFO天天盯着。
很多人觉得人家吹牛逼其实东西做得不好,那是根本上没有理解自己工作的内容和目的。
哈哈
mark 一下这个
不考technical题目的话招进来的人都比较坑爹,这是你老板的问题。我自己面了几家公司,都考得特别仔细,考leetcode题目,如果做ml的就问模型,手推公式,如果做ab test就问怎么设计实验怎么测实验结果,最不行也会给题目让你在家做交回去...