一波挂经

一波挂经 难受
Factual: (glassdoor) Associate Data Scientist 这家公司人都超级的Nice, 面试聊天非常开心,虽然挂了但是感觉是个很好的公司
HR: experience, background
Take Home, ugly json data set filtering and cleaning, pandas not working here
1st tech: coding: split address phone number, what attribute do you expect to see in geolocation(business sites) dataset, how do you know if a new dataset add value to the existing data set in Factual
2nd tech: questions about a past project and challenges in the project, coding: balance bracket, how do you design a classifier to classify if the weather is good in a city(where do you collect data, how to label, how do deal with people in different region have different standard for good weather) - 挂, excellent coding, needs to improve critical thinking case 做的不好

小精灵: (refer) Data Scientist 打着了解product的名号 玩了好几天游戏 罪过
Phone Screen, SQL, Case(How do you evaluate two campaigns in France, one is on the 10th of July and the other one is on the 15th) - 挂, SQL写的磕磕绊绊然后还出了基本的错误然后面试官脸就黑了(那时候就知道挂了),总体很偏business not the best fit

Plume Design: (glassdoor) Data Scientist 面试官都是天竺人,办公室非常的简陋,屋子里一股午饭味
Phone Tech: porbs that 3 or more coins are head if you filp 5 coins at the same time, tell me about a sport you like and design a machine learning algorithms that updates the winning probabilities for each side (choose snooker, described rules and metrics, some Bayesian model, follow up Markov model briefly described)
Onsite: NDA 4轮,每个人都问了3个问题, 有coding(< medium), ML基础概念(clustering, PCA, regression, time series, recommender system) SQL(easy), 还有给case让设计abnormality detection algorithm given a limit number of read operations from multiple machines- 默据,project 讲的不太好,给老大讲SARIMAX还有project的细节没讲明白,ML有的基本概念讲的不好

ZestFinance: (glassdoor) Data Scientist
HR Phone
OA 2hr limit classification with unbalance data
Phone Tech: Question about recent project,how is the feature importance calculated in tree models - 挂 迷

Vectra: (glassdoor) Data Scientist
HR Phone: past project, interest - 挂 迷

Petuum: (refer) Software Engineer - Machine Learning 被吊打 心服口服
Phone Tech 1 , question about past project, neural PCA, define loss function for PCA and write it with TensorFlow, 没写出来loss,一直在想covariance matrix, 实际是在考PCA的推倒应该min L2 error 以为挂了
Phone Tech 2 (round 1 一个多月后), What is PCA, write Hinge loss function, compare PCA and factor analysis, write batch normalization 挂 以为要考coding 刷了几天题, 一点没复习DL 哎~

积攒人品 也希望能帮到大家 祝好运

请问可以私信聊一聊zestfinance data challenge吗

就是给非常好的data然后有时间,信用分数什么的 做一个是不是approve的model 然后选个model evaluating metric(auc, f1 score etc.) 再说一下原因,给了一个test_x的csv, 把你model predict的结果csv还有一个小report发回去就行了

谢谢分享! data是不是不用做什么处理呢基本上 直接建模?

差不多 时间可能要稍微处理一下