It's Data Science Jay here and today I want to talk about some common behavioral interview questions for data scientists and machine learning engineers and how we can answer them.

我是數據科學 Jay，今天我想談談數據科學家和機器學習工程師常見的一些行為面試問題，以及我們該如何回答這些問題。

Mainly I want to go over frameworks and common strategies that we can use to solve the behavioral interview questions.

我主要想介紹一下我們可以用來解決行為面試問題的框架和常用策略。

Also if you're new here, my name is Jay.

另外，如果你是新來的，我叫傑伊。

I'm a remote entrepreneur working in the data science space that used to be a data scientist.

我是一名在數據科學領域工作的遠程企業家，曾經是一名數據科學家。

Now I've founded a startup called Interview Query which is the premier data science interview platform.

現在，我創辦了一家名為 Interview Query 的初創公司，它是首屈一指的數據科學麵試平臺。

So check it out but if you're new here please like and subscribe as well.

如果你是新來的，也請點贊和訂閱。

So for behavioral interviews there's a lot of videos out there about how to answer them and how to actually be good at them and those are all great but how do we actually communicate them specifically for data scientists and machine learning engineers or anyone else working in data.

是以，對於行為面試，有很多關於如何回答面試問題以及如何真正做好面試的視頻，這些都很好，但我們如何針對數據科學家、機器學習工程師或其他從事數據工作的人實際傳達這些資訊呢？

Generally I see it kind of conform into three main buckets.

一般來說，我認為主要分為三類。

The first one is about how well you can actually communicate technical concepts when you're answering behavioral interview questions.

第一個問題是，在回答行為面試問題時，你究竟能不能很好地傳達技術概念。

The second one is figuring out how to actually back up your resume and make sure that it holds up under questioning.

第二個問題是如何為履歷提供實際支持，確保履歷在質疑聲中站得住腳。

And then the third kind of behavioral interview question I see is more around culture fits and making sure that you're a good fit for the team.

我看到的第三種行為面試問題更多是圍繞文化契合度，確保你適合團隊。

So generally when people think about behavioral interview questions they kind of think about the recruiter screening right.

是以，一般人在考慮行為面試問題時，都會想到招聘人員的篩選。

A recruiter will call you up, they'll ask you a little bit about yourself, some projects that you've done, your visa sponsorship, where you're located and stuff like that.

招聘人員會打電話給你，問你一些關於你自己的情況、你做過的一些項目、你的簽證擔保、你在哪裡等等。

And so in general I don't think that's how most behavioral interview questions that actually matter but those aren't really the behavioral interview questions that matter.

是以，總的來說，我不認為這就是最重要的行為面試問題，但這些並不是真正重要的行為面試問題。

More common than ever hiring managers and executives are actually the ones that are screening data scientists and engineers up front in the first one or two interviews.

實際上，招聘經理和高管在前一兩次面試中篩選數據科學家和工程師的情況比以往任何時候都要普遍。

The reason why that's happening is because a lot of the hiring managers and executives figure out it's basically a waste of time if you pass all the technical interviews but then on the last behavioral interview you fail because of the fact that you have to pass all the interviews to actually get the offer.

之所以會出現這種情況，是因為很多招聘經理和主管認為，如果你通過了所有的技術面試，但在最後一次行為面試時卻失敗了，這基本上是在浪費時間，因為事實上，你必須通過所有的面試才能真正拿到錄取通知。

Because the executives have more judgment and more control over the hiring process many times they'll actually ask the behavioral interview questions at the very front to then reject you or move you on.

因為高管對招聘過程有更多的判斷和控制，很多時候他們會在最前面提出行為面試問題，然後拒絕你或讓你繼續工作。

The key with demonstrating technical competency in the behavioral interview questions is to have two things.

在行為面試問題中展示技術能力的關鍵在於兩點。

One is strong technical skills and then the other is strong behavioral interview and communication skills.

一個是過硬的技術技能，另一個是過硬的行為面試和溝通技能。

You could have all the technical knowledge in the world but if you can't communicate it then it's basically useless to everyone else.

你可以擁有世界上所有的技術知識，但如果你不能將其傳播出去，那麼這些知識對其他人來說基本上是無用的。

So a key thing to practice around technical communication is understanding exactly how to craft your stories for different kinds of interview questions.

是以，圍繞技術交流進行練習的關鍵是瞭解如何針對不同類型的面試問題準確地編寫自己的故事。

Let's go over an example of when I was actually a hiring manager once for a prior company and asked an intern a behavioral interview question that they didn't really pass.

讓我們舉個例子，我曾經在一家公司擔任招聘經理，向實習生提出了一個行為面試問題，但他們並沒有通過。

It was actually a pretty simple question.

這其實是個很簡單的問題。

I asked the data science intern candidate, tell me about a project that you're proud of and then the candidate responded with a lot of enthusiasm but they responded with an answer like this.

我問數據科學實習生候選人，告訴我一個你引以為豪的項目，然後候選人滿懷激情地回答，但他們的回答是這樣的。

Yeah for my CS 100 class we made these robots and we programmed them using java and my robot won the competition and the TA said it was really good.

是的，在我的 CS 100 課上，我們做了這些機器人，用 java 編程，我的機器人贏得了比賽，助教說它真的很棒。

Okay so how did you end up building it?

好吧，你最後是怎麼造出來的？

We used java.

我們用的是 Java。

Okay but what actually made the robot good?

好吧，但機器人究竟好在哪裡？

Well it won the competition.

它贏得了比賽。

You'd actually be surprised by how many of those answers that you get when you're actually hiring for data candidates and it's not that the project itself was bad it's just the fact that the candidate didn't go into any details that makes the project worth describing and as an interviewer everything we really care about is in the details of the project.

實際上，在招聘數據候選人時，你會驚訝地發現有很多這樣的答案，這並不是說項目本身不好，而是候選人沒有詳細說明項目值得描述的細節，而作為面試官，我們真正關心的是項目的細節。

I mean obviously there's a fine line here I don't want you to get super granular telling me about the functions you use and the code base and stuff like that but generally it's about demonstrating that you actually did the project that you have the technical expertise to do it again and the fact that you use strategies and different kinds of frameworks to help you navigate through the project while working with other team members.

我的意思是，這裡顯然有一個細微的界限，我不希望你告訴我你使用的函數和代碼庫之類的東西，但一般來說，這是要證明你確實做了這個項目，你有再次做這個項目的技術專長，以及你在與其他團隊成員合作時，使用策略和不同類型的框架來幫助你完成項目。

The best answers to behavioral interview questions like these are to frame it in a story and I know we have all these different kinds of frameworks like STAR to actually go through the situation, task, action, result and those are great to use but let's go through an example of how I'd actually answer this interview question if I was asked it.

我知道我們有各種不同的框架，比如 STAR 框架，可以通過情況、任務、行動、結果來進行實際操作，這些都很好用，但讓我們舉例說明，如果有人問我這個面試問題，我會如何回答。

Tell me about a time that you use data science to make an impact on the business.

請告訴我您利用數據科學對業務產生影響的案例。

That's a great question.

這個問題問得好。

So at my company called Jobber basically it was a tinder style swiping app so you would swipe right to apply to jobs and you swipe left pass on jobs and my job as a data scientist was to improve the recommendation engine.

我所在的公司叫 Jobber，基本上是一款 "Tinder "式的刷屏應用，你可以向右刷屏來申請工作，向左刷屏來放棄工作，而我作為數據科學家的工作就是改進推薦引擎。

The existing model that was used in the recommendation engine before was this naive bayes trained model and it worked all right but I felt like there could have been improvements to it because of the fact that we were getting a lot of customer complaints about the model recommendations when they would upload their resume and they get a bunch of jobs that didn't actually match up to what they were looking for.

之前推薦引擎中使用的現有模型是天真貝葉斯訓練模型，它的效果還不錯，但我覺得它還可以改進，因為我們收到了很多客戶對模型推薦的投訴，當他們上傳履歷時，他們得到的一堆工作實際上與他們正在尋找的工作並不匹配。

So instead I built an elastic search model that actually took a lot of the synonyms from their resume and then added it to the elastic search query parameter to return better jobs and also more jobs so they wouldn't actually run out of jobs from the previous model.

是以，我建立了一個彈性搜索模型，實際上是從他們的履歷中提取了很多同義詞，然後將其添加到彈性搜索查詢參數中，以返回更好的工作和更多的工作，這樣他們就不會因為之前的模型而找不到工作了。

So the way that we actually tested this I didn't actually know if it was going to be better or not so we launched an A-B test we put only 10% of the users into the new elastic search model and then analyzed the total number of applications for both the control group and the experiment group.

是以，我們實際測試的方式是，我不知道這樣做是否會更好，所以我們進行了一次A-B測試，只將10%的用戶放入新的彈性搜索模型中，然後分析對照組和實驗組的應用總數。

Turns out that the new elastic model actually improved job recommendations and improved the amount of total applications by about 20% which in turn increased the amount of revenue that we made on the top line by around 10 to 15% because we get paid per application.

事實證明，新的彈性模式實際上改進了職位推薦，並將申請總量提高了約 20%，這反過來又使我們的收入增加了約 10%至 15%，因為我們按申請獲得報酬。

So quickly just to deconstruct that answer notice how in the very beginning I stated the exact problem that was going on.

請注意，我在一開始就準確地指出了問題所在。

We needed a new recommendation engine and that was my job to solve it.

我們需要一個新的推薦引擎，我的工作就是解決這個問題。

Then I went into the details on exactly how I solved it by building this new elastic search query engine and at the very end I just detailed the results of what happened and how we launched it and what the business impact actually gave.

然後，我詳細介紹了我是如何通過建立這個新的彈性搜索查詢引擎來解決這個問題的，最後，我詳細介紹了所發生的事情的結果、我們是如何啟動它的，以及它對業務的實際影響。

So it's not that hard it's more about crafting the story in a way in which you kind of define what the problem is and kind of explain and go through the details.

是以，這並不難，更多的是以一種定義問題的方式來編寫故事，並對細節進行解釋和說明。

Notice how I did kind of go into details about the different query parameters we and how it was naive Bayes.

請注意我是如何詳細介紹我們不同的查詢參數以及如何使用天真貝葉斯的。

I dropped those as buzzwords because I know data science and I'm showcasing that I know data science.

我放棄了這些流行語，因為我瞭解數據科學，我正在展示我對數據科學的瞭解。

But at the end of the day what matters is the fact that you can showcase that business impact you can talk it through and you can communicate that to someone else just like how you communicated that to your manager when you're first working on the project.

但歸根結底，最重要的是你能展示業務影響，你能把它說清楚，你能把它傳達給別人，就像你剛開始做項目時如何把它傳達給你的經理一樣。

Adding structure to any sort of behavioral interview question answer is a huge win because one it just makes your story more engaging.

在任何一種行為面試問題的回答中加入結構都是一個巨大的勝利，因為它可以讓你的故事更吸引人。

A lot of times I hear these stories from candidates and just honestly my mind goes elsewhere because you know I have better things to do.

很多時候，我從候選人那裡聽到這些故事，說實話，我的心思都跑到別的地方去了，因為你知道我有更好的事情要做。

I was hiring data scientists but if you make your actual story engaging if you actually try to add some structure to it your behavioral interview question answers will go a lot smoother.

我當時正在招聘數據科學家，但如果你能讓你的實際故事引人入勝，如果你能真正嘗試為故事添加一些結構，你的行為面試問題回答就會順利得多。

A couple more behavioral interview questions are really common to data scientists.

還有幾個行為面試問題對數據科學家來說確實很常見。

I want to run through these really quickly.

我想快速瀏覽一下。

Tell me about a time that you had to clean and organize a big data set.

請告訴我，你曾有過清理和整理大型數據集的經歷。

How have you used data to elevate the experience of a customer or stakeholder.

您是如何利用數據提升客戶或利益相關者的體驗的。

Tell me about a time you had to communicate something technical to a non-technical person and then how would you communicate data driven insights to a business stakeholder.

請告訴我，有一次你不得不向非技術人員傳達一些技術性的東西，然後你會如何向業務利益相關者傳達數據驅動的見解。

Next up the second most important thing that I think happens in these behavioral interview questions is analyzing your resume and talking through your resume.

接下來，我認為在這些行為面試問題中，第二件最重要的事情就是分析你的履歷，並通過履歷進行交流。

So a lot of time I'm pretty critical of someone's resume especially because of the way that anyone can kind of add anything to their resume without a lot of backing to it and so one of the key things I do as a data science hiring manager was is to basically go through and actually ask very specific key points about their resume that they listed below.

是以，很多時候，我對履歷很挑剔，特別是因為任何人都可以在履歷中添加任何內容，而不需要很多支持，所以作為數據科學招聘經理，我所做的關鍵事情之一就是，基本上要通過他們的履歷，並實際詢問他們下面列出的關於履歷的非常具體的關鍵點。

For example once I was asked to interview a candidate who said on their resume that they had increased the company baseline revenue by 30% month over month.

例如，有一次我被要求面試一位應聘者，他在履歷中說自己的公司基線收入比上個月增加了 30%。

I was pretty skeptical about this but I kept on asking questions and turns out the business was an e-commerce business and the month over month change was from November to December which as you know is Christmas buying shopping season.

我對此非常懷疑，但我繼續追問，結果發現這家公司是一家電子商務公司，月度變化是從 11 月到 12 月，而大家都知道 12 月是聖誕購物季。

And so when I asked them what happened in the month of January suddenly they didn't have as much to say about their project influencing revenue.

是以，當我問他們一月份發生了什麼事時，他們突然對影響收入的項目沒有那麼多話可說了。

So the core kind of learning here is really just to back up your resume and have a story and make sure that it's sound proof and tight for every single point that you put on your resume.

是以，學習的核心內容就是為你的履歷提供支持，編寫一個故事，並確保履歷上的每一個要點都有理有據、嚴絲合縫。

Generally I limit my resume bullet points to around four core bullet points for like the four biggest projects that I contributed to for every single job in my past career.

一般來說，我會把履歷要點限制在四個核心要點左右，比如我在過去職業生涯中每份工作都參與的四個最大的項目。

If you work in an analytics capacity where you're very business focused make sure you're not inflating a lot of those numbers or at least that you have a way to back exactly what you contributed to within analytics.

如果你從事的分析工作非常注重業務，請確保你沒有誇大這些數字，或者至少你有辦法確切地證明你在分析中的貢獻。

Similarly if you're in a machine learning role kind of focus make sure that you know exactly the technical details of the machine learning projects or the technical projects that you implemented.

同樣，如果你從事的是機器學習方面的工作，也要確保你對機器學習項目或你實施的技術項目的技術細節瞭如指掌。

If you put neural networks in your resume or something about PyTorch and I know a lot about PyTorch then I'm going to ask you a lot about it and expect that you have that same level of knowledge.

如果你在履歷中提到神經網絡或 PyTorch，而我對 PyTorch 非常瞭解，那麼我就會問你很多關於它的問題，並希望你也有同樣的知識水準。

Generally if you can't explain your resume very well to an interviewer that's a huge red flag and probably one of the biggest red flags there are because then people think you might be lying about other things as well.

一般來說，如果你不能很好地向面試官解釋你的履歷，這就是一個巨大的紅旗，也可能是最大的紅旗之一，因為這樣人們就會認為你可能在其他事情上也在撒謊。

So in general just make sure that everything on your resume you can actually back up.

是以，一般來說，只要確保履歷上的所有內容都能得到實際支持就可以了。

Lastly a lot of these behavioral interview questions for data scientists come around and culture fit is a huge issue for technical candidates because a lot of the times you get these really really strong technical candidates but that are really really bad at communication or working well with others and so generally for most of these culture fit questions they're testing you on a couple things.

最後，很多針對數據科學家的行為面試問題都會出現，對於技術候選人來說，文化契合是一個巨大的問題，因為很多時候，你會遇到一些技術能力非常強的候選人，但在溝通或與他人合作方面卻非常糟糕，所以一般來說，對於大多數文化契合問題，他們都會在幾個方面對你進行測試。

One is straight up that you're not an asshole they want to know that you can work well with others.

一是直接說你不是個混蛋，他們想知道你是否能與他人很好地合作。

If you're a senior candidate they want to know that you can actually mentor junior team members on the same team and then three if you're more of a junior candidate they want to know that you're eager to learn and you can demonstrate initiative for getting better in your role.

如果你是資深應聘者，他們想知道你是否能指導同一團隊中的初級團隊成員；如果你是初級應聘者，他們想知道你是否渴望學習，是否能在自己的崗位上表現出不斷進步的主動性。

Most of the time culture fit is really hard to evaluate and there's not one size fit all in terms of how you can answer these questions.

大多數情況下，文化契合度真的很難評估，而且在如何回答這些問題方面，也沒有一個放之四海而皆準的方法。

The sad reality of it is that a lot of these companies have different cultural characteristics for how they run their companies and this is important for you as a candidate also to figure out because it might not be the same cultural fit that you're aligned with.

可悲的現實是，很多這樣的公司在如何經營自己的公司方面有著不同的文化特點，這對作為候選人的你來說也很重要，因為這可能與你的文化不一致。

For example one company might really really enjoy cross-functional communication and also a lot of meetings but if you're the person who doesn't like meetings and really likes to focus on the code base or pumping out good code then it might not be a great cultural fit and you should also excuse yourself from the interview process then as well.

例如，一家公司可能非常非常喜歡跨職能溝通，也非常喜歡開很多會議，但如果你是一個不喜歡開會、喜歡專注於代碼庫或編寫優秀代碼的人，那麼這家公司可能與你的文化不太契合，你也應該放棄面試。

Probably the most common interview questions that come with this are more so around the situational kind of star framework type of questions where you can talk about situations where you demonstrated leadership or initiative to really make yourself seem like a really good cultural fit and team member.

最常見的面試問題可能更多是圍繞情境類的明星框架型問題，你可以談談你在哪些情況下表現出了領導力或主動性，從而真正讓自己看起來像一個非常適合文化和團隊的成員。

So examples are tell me about a data project that you've worked on where you encountered a challenging problem, how did you respond, how have you gone above and beyond the call of duty, tell me about a time that you failed and what you learned from it, how did you handle meeting a tight deadline, tell me about a time when you resolved a conflict, provide an example of a goal you reached and tell me how you achieved it.

舉例說明，在你參與的一個數據項目中，你遇到了一個具有挑戰性的問題，你是如何應對的；你是如何超越職責要求的；告訴我你有一次失敗了，你從中學到了什麼；你是如何應對緊迫的最後期限的；告訴我你有一次解決了衝突；舉例說明你達到了一個目標，並告訴我你是如何實現的。

So in summary make sure that one you can communicate technical concepts and provide good stories and good structure for the projects you've worked on and also the basis for your technical proficiency.

是以，總而言之，要確保自己能夠傳達技術概念，為自己參與過的項目提供好的故事和好的結構，這也是自己技術能力的基礎。

Number two is to make sure to back up your resume and make sure it doesn't have any red flags on it and number three is just to overall assess culture fit in the best way you can demonstrate that you can show initiative and be a good team member with other people on the team.

第二，確保你的履歷有充足的證據，確保履歷上沒有任何 "紅旗"；第三，以最佳的方式全面評估文化契合度，證明你能展現出主動性，並能與團隊中的其他人一起成為優秀的團隊成員。

Lastly I want to talk about today's sponsor which is Interview Query.

最後，我想談談今天的贊助商，它就是 Interview Query。

Interview Query is a data science interview platform that I am the founder of.

Interview Query 是一個數據科學麵試平臺，我是該平臺的創始人。

Basically we have hundreds of interview questions, courses, sequel editors, challenge assessments so you can assess your ability against other people.

基本上，我們有數以百計的面試問題、課程、續集編輯、挑戰評估，這樣您就可以對照其他人來評估自己的能力。

It is probably the best technical data science interview platform out there and while we're about to launch behavioral interview questions on the site we haven't yet but if you're getting ready for a technical interview be sure to check us out.

它可能是最好的技術數據科學麵試平臺，雖然我們即將在網站上推出行為面試問題，但我們還沒有推出，但如果你準備參加技術面試，請一定來看看我們。

Thanks for watching and I'll see you all later.

感謝觀看，我們稍後再見。

Bye.

再見。