Q&A with a Data Science Hiring Manager | Coursera Community
Coursera Header
Q&A

Q&A with a Data Science Hiring Manager

  • 27 February 2019
  • 46 replies
  • 2030 views
Q&A with a Data Science Hiring Manager

Show first post
This topic has been closed for comments

46 replies

Userlevel 2
Badge
So many good questions! :)

I need to head into work, but I look forward to answering more tomorrow and the rest next week!
Badge
Hi Dan,

Thank you for sharing you opinion on the data scientist role. Would really appreciate it if you can kindly answer some questions that are haunting me for a while.

  • How do you distinguish some must-have skills from good-to-know skills for a data scientist role? The data science is a very interdisciplinary subject. As there are so many skills are necessary to learn, how do you choose learn one first over another?
  • I noticed that you have an interesting experience in Zipfian Academy's program. How do you think of this kind of bootcamp experience in your career?
  • Do you think it is necessary to read lots of papers to become a real data scientist? If so, how did you teach yourself reading these papers and applying them to your job?
  • How do you think of a startup work experience when you are interviewing for Coursera? Skills used every day in a startup may be very different from skills learned from academia. Do you prefer a professional with years in startup or a new graduate from a Phd program?
Badge
Hi Dan,

Thanks for your time and effort!

I have few questions how to merge data science and business:

  • How you translate the output of the data processor / model into business insights?
  • How seamless is this transition?
  • To what extent human aspect play the role (e.g. management does not understand math behind algorithms etc)?
  • How you overcome those challenges (trainings, use cases etc)?
Thanks a lot,
Best regards,
Misha
Userlevel 2
Badge
What do you expect from a recent grad when you ask questions like - where do you see yourself in 2 years from now ? I have been asked this question recently during an onsite interview for a start up.

Thank you for the question!

My general advice here is to tell a compelling story: People love digestible narratives. More importantly, they love hearing how they fit into your narrative :)

Concretely, what I mean is to explain how the opportunity you're interviewing for fits into your broader career arc, e.g., "I've done X and Y [in school/internships/personal projects/etc.]. I think I would like to become Z, and this opportunity excites me because I think it will help me develop the skills I need to become Z."

Of course, don't fear if you're not sure where you want to be in the future! I think most people are generally sympathetic to the idea that the future is intrinsically uncertain, especially for a new graduate :)

(FWIW, I don't typically ask that question in interviews, but if I did, it would be because I want to ensure that the role you're interviewing for could help you accomplish your career goals. I'd hate for you to have a well-defined career goal that simply wouldn't be well-served by the opportunities I have available.)
Userlevel 2
Badge
  • While hiring Data Scientist how to differentiate the role communicated, e.g. ML engineer vs Deep Learning Engineer vs AI Algorithm developer vs Data Scientists doing Analytics/visualization/reporting?
  • Do you expect them to solve puzzles/tower of honoi/search/sort type of problems during interview in addition to all the ML/DL/Stats thing?


Thank you for the questions! Answers below:

  • While hiring Data Scientist how to differentiate the role communicated, e.g. ML engineer vs Deep Learning Engineer vs AI Algorithm developer vs Data Scientists doing Analytics/visualization/reporting?


At Coursera, we try to ensure that we're very clear about what the responsibilities of any given role are. To this end, we break down our Data Team into three main functions: Data Engineering, Data Products, and Decision Science.

  • Data Engineering builds our core data warehouse and maintains our self-serve BI tooling (e.g., Looker, Tableau).
  • Data Products does "Data Science for Machines," i.e., they build models that power some kind of onsite application (e.g., they build recommendation engines, algorithms that improve search, etc.)
  • Decision Science does "Data Science for Humans," i.e., they perform analyses and build models that help business or product stakeholders make important decisions (e.g., they analyze experiments, suggest productive improvements, analyze the impact of X feature among Y group of users, etc.)
The output of Data Products work is usually some kind of API that product engineers can call to personalize the onsite experience. The output of Decision Science work is usually some kind of Jupyter Notebook or presentation that makes a strategic recommendation.

Of course, this isn't the only way of separating Data Science roles. Airbnb, for example, splits their team into Analytics, Inference, and Algorithms. (At Coursera, we combine "Analytics" and "Inference" into "Decision Science," with the expectation that our cross-functional partners can do most of their own analytics through our self-serve BI tools mentioned above.)

My advice here would be to ask precisely what a given company has in mind when they say "Data Scientist." Ask, "Am I going to be building models that will be put in production?" "Am I going to be analyzing data to help Product Managers or Marketers make decisions?" "Am I going to be need engineering chops, or are you really looking for a statistician?"

Unfortunately, job postings can often be very unhelpful in distinguishing among these roles. That said, there are usually keywords that signal whether a given opportunity is Data Products or Decision Science. (Data Products: Machine Learning, Algorithms, Engineering, etc. Decision Science: Analytics, A/B Testing, Statistics, etc.)

  • Do you expect them to solve puzzles/tower of honoi/search/sort type of problems during interview in addition to all the ML/DL/Stats thing?


We don't, and FWIW, I didn't encounter these types of problems during my job hunt. (That said, I wasn't interviewing for ML Engineering roles, and it's possible they'd be more common in that realm.)

Hope this was helpful -- thanks again!
Badge
How well do people with Coursera credentials fare vis-à-vis people coming out of traditional colleges and universities with Bachelors, Masters or PhD degrees?
Userlevel 7
Badge +13

Thanks for all your questions! This thread is closed to new questions.

Userlevel 2
Badge
Q1 : What platforms and tools does the coursera Data Science team use to build and deploy models ?

Q2 : Do you have a prize such as 'Data Science' Oscar for students who have completed maximum count of courses from various universities related to Data Science on your platform?

Q3 : What's the maximum amount of time spent on the platform and the count of courses completed by a single coursera student ?


Thank you for the questions, Addy! Taking them in order:

Q1. In terms of prototyping models, we generally use a combination of R and Python. When it comes time to deploying our models, we have some wonderful in-house tooling built by our Data Infrastructure team :)

Q2. I don't think we do, but that's a great idea! I'll pass it along to our Marketing team :)

Q3. I'm not positive I can answer this; however, we have a ton of very dedicated learners who've completed quite literally dozens and dozens of courses.
Userlevel 2
Badge
I've been analyzing wind turbine noise data for about 2 years. This is a very specialized field therefore the number of job openings in this area is also limited. When I apply to data scientist positions from other domains (like gaming for example) I get rejected indicating the lack of my domain knowledge.

How can I convince a hiring manager that I can learn about the domain (although it's new to me) and be effective in the job?


Great question, Thileepan!

I would lean on some advice I already gave (e.g., develop a DS portfolio, etc.). However, in your case, you're lucky in that your experience is already analytical.

I'm not necessarily sure if this applies to you, but one thing I've noticed in folks transitioning from more "niche"/specialized fields to Data Science is that their resume/cover letters/etc. sound too much like they're optimizing for their "past" career, not the career they want.

To put it bluntly, hiring managers aren't always the most creative people. They're trying to find someone they think can do the job as quickly as possible, and if a resume or cover letter requires them to think creatively (e.g., "Well, this candidate doesn't have experience in my domain, but their field *sounds* technical, so maybe it's worth a shot..."), it's going to reduce your changes of getting called back.

In my case, for example, when I was transitioning from Finance, I learned that my resume was too "finance"-y. In particular, it talked a lot about backtesting algorithms on historical S&P 500 data, and tech companies didn't care about backtesting algorithms on historical S&P 500 data. However, they might have cared about my ability to build and deploy models that brought value to our cross-functional stakeholders.

In short, a job application is an opportunity to tell a story, and when you're transitioning from one field to another, the story should be: "Here's how my experience uniquely prepares me for helping you solve the challenges you're facing." People are usually amenable to compelling narratives, and that's a compelling narrative.

(One additional note: to pull this off, it's important to develop a fluency with the jargon of the company/industry you're pursuing. Networking is a great way of learning that jargon 🙂
Userlevel 2
Badge
Q1: How would you draw the boundary between Data Analyst, Data Engineer and Data Scientist roles?
Q2: Software development has methodologies like Agile, Scrumm and etc. They are not easily applicable to Data Science because of blurry timeframe. What do you employ?


Thanks for the questions, Wladd!

I tackled Q1 in another response; however, for Q2, I would point you toward my favorite blog post on this topic. In it, the author builds the case for a "modified" Agile methodology that helps balance the longer-term "research"-y work with executing on short-term deliverables.

I've found that it works quite well, although the exact implementation would probably differ from one team to the next, or even from one quarter to the next (e.g., in some months, we want to focus more on exploratory work; at other times, we need to be deliver on shorter-term initiatives).
Userlevel 2
Badge
Hi. I am currently studying post grad Data Science and have had some of the Coursera courses related to it. But before deciding to explore Data Science, I was a software developer for a number of years.

Questions:
  • What is the best way to marry and leverage both Data Science skills and Software Development skills in the current job market?
  • Talking with a number of employers at career fairs, they either just want a developer OR a data scientist. Are there opportunities in the job market that builds on both of the skills or is it usually just one or the other?


Thank you for the thoughtful question!

The short answer is: YES, absolutely. The most accomplished Data Scientists I know tend to have stronger-than-usual software engineering skills. Indeed, the ability to "think like a Data Scientist, but build like an engineer" makes you uniquely valuable. Whether it's deploying machine learning models in production, making your analysis more reproducible, or figuring out ways to improve your ETL processes, a software engineering toolkit allows you to amplify your Data Science efforts in really compelling ways.

I would probably give the following advice:

(a) Pursue opportunities in ML Engineering (assuming you're interested in this part of Data Science): More and more companies are combining "model builders" and "model deployers" into a single role: The ML Engineer. Given your background, this could be a good fit.

(b) Focus on smaller companies: At larger, more established companies, there are going to be teams that only do Data Engineering, teams that only do Analysis, teams that only build models, etc. As such, having a multidisciplinary skill set might be less valuable than at a smaller company, which literally doesn't have the resources to hire a research person separate from a Data Engineer.

Hope this is helpful, and best of luck!
Userlevel 2
Badge
How could a data scientist improve himself on business problem-solving?

There are some good material/courses about applying data science to solve business problems? One can know all about the algorithms, the features, the training of a model, but being able to solve the actual company problem and turn these into profit is something else.

Thanks and my best regards.


Thank you for the thoughtful question, Danilo!

I think your point is right on: Thinking about your work in terms of business impact is absolutely critical. If you have that lens, you'll always be a valuable team member. (The number of Data Scientists who optimize for what's technically interesting is surprisingly high.)

In general, I'm not sure if there are any shortcuts here (or if there, I haven't found any :)

My approach/general principle has been learning as much as I can about the domains of the people I'm partnering with.

For example, at one point, I began working with our Marketing team more and more often. To better understand their world, I took Coursera's Digital Marketing Specialization. Learning what problems Marketers cared about enabled me to apply my Data Science skill set to their problems (not what I thought their problems were), which enabled me to bring a ton of value. (In particular, as I was going through the course, I kept thinking, "Oh, Data Science can be useful here, here, and here," or, "Wow, we could do these kinds of analyses easily.")

Similarly, when I began working with the Growth Team, I learned as much as I could about Consumer Growth by reading the blog posts of Andrew Chen, along with every book on growth hacking I could find on Amazon.

Boiling it down to some general principle, I'd say this: As Data Scientists, we have a ton of unique skills, but they're only valuable when they're intersected with some kind of domain knowledge. The trick, then, is building that domain knowledge. As you do, problems to solve will naturally emerge.
Userlevel 2
Badge
Hello dan,
My question is how do you judge approach of a candidate, in solving a data science problem, and according to you what are the most common approaches preferred by interviewer.


Thank you for the question!

I think the single most impressive thing candidates can do to stand out is be clear about their assumptions. A lot of interviewees leap immediately to solving the problem, but it's impressive when someone can step back and say, "Well, I think I would approach the problem this way. However, it's contingent on X or Y assumption. If it actually turned out that X and Y weren't true, then I think I'd want to do things a little differently."

This shows a level of analytical flexibility, creativity, and rigor that is the hallmark of a good Data Scientist!
Userlevel 2
Badge
  • How can a software engineer working on databases and programming on a daily basis prepare for a career in data science ?
  • What are some platforms(apart from kaggle) where we can showcase our data science interest and personal work to data science hiring managers ?


Thank you for the questions, Prajna! Taking them in order:

  • How can a software engineer working on databases and programming on a daily basis prepare for a career in data science ?


One of my favorite bits of advice is: Look for ways to be a "Data Scientist" in your current job :)

Concretely, when I was working in finance, I explained to my manager how certain Data Science techniques I was learning on Coursera could be used to solve our business problems. It wasn't easy to convince him, but once I did, I suddenly had "formal" permission to get some Data Science experience.

In the same way, I've seen Engineers at Coursera who learned how to analyze simple experiments or leverage our self-serve BI tools to answer questions about how well their products were working. Their pitch was: "Hey, I'm closest to the problem. If I can learn how to analyze what's working and not working, I can be more agile in making improvements."

  • What are some platforms(apart from kaggle) where we can showcase our data science interest and personal work to data science hiring managers ?


I wouldn't underestimate the power of a humble blog! When I was transitioning from finance to Data Science, I created a blog (dsaber.com) that showed off some of my projects and ability (most of my blog posts linked to code on GitHub). The fact that I could demonstrate my communication and technical skills -- and speak to my projects in interviews -- helped me land my first job.
Userlevel 2
Badge
  • Do you use take home challenges in hiring? If so, what do you look for? What are your general thoughts on take home challenges?
  • What has changed for you in going from a data scientist to a data science manager? What are the skills you needed to demonstrate to make that jump?
  • What is the split in how much of your team's work is on doing analytics looking at the past and predictive analytics?


Thanks for the questions! Answers below:

  • Do you use take home challenges in hiring? If so, what do you look for? What are your general thoughts on take home challenges?


We do use take-home challenges. As with anything, they have pros and cons:

Pros:
  • Ideally, they're a more unbiased gauge of someone's ability
  • They allow us to "interview" many more people -- if we had to rely 100% on resumes and phone screens, we would end up screening out many qualified candidates
  • Tangentially, they allow folks without experience to "prove" themselves more directly
Cons:
  • Some take-home challenges I've seen can take a long time. As such, they might be harder for folks with a lot of commitments outside of work. (Note that ours only takes 90 minutes!)
  • They're not a completely "true" guide to how someone would fare at work: after all, how many multiple choice questions do you answer in the course of your work day?
In our case, we thought the pros outweighed the cons, especially insofar as it allows us to cast a wider net and tap into a much larger pool of applicants that don't necessarily fit the "traditional" Data Scientist profile. (Of course, the efficacy of this approach depends on how thoughtfully it's implemented.)

  • What has changed for you in going from a data scientist to a data science manager? What are the skills you needed to demonstrate to make that jump?


I think this question would really be worthy of an entire essay (which I will one day hopefully write :)

For the time being, I would say that the single biggest change was this: I had to shift my mindset from, "How do I solve this specific problem?" to "How I do set up an environment where members of the team can optimally solve their specific problems?"

Ultimately, as a manager, your output doesn't matter. Only the output of your team does. That is an admittedly obvious statement, but it's a mind-bogglingly difficult one to internalize, and I still work on doing so each and every day.

In terms of what skills you'd need to demonstrate, I would highlight three: humility, empathy, and the ability to think globally rather than locally.

  • What is the split in how much of your team's work is on doing analytics looking at the past and predictive analytics?


The short answer: It depends :)

In some quarters, I'd say the majority of our work is analytics, and in other quarters, it flips. Indeed, in most cases, the analytics ends up fueling the predictive efforts, so theoretically, it should be a virtuous cycle.

Thank you again for the great questions! Hope this was helpful 🙂
Userlevel 2
Badge
Hi Dan,

Thank you for sharing you opinion on the data scientist role. Would really appreciate it if you can kindly answer some questions that are haunting me for a while.

  • How do you distinguish some must-have skills from good-to-know skills for a data scientist role? The data science is a very interdisciplinary subject. As there are so many skills are necessary to learn, how do you choose learn one first over another?
  • I noticed that you have an interesting experience in Zipfian Academy's program. How do you think of this kind of bootcamp experience in your career?
  • Do you think it is necessary to read lots of papers to become a real data scientist? If so, how did you teach yourself reading these papers and applying them to your job?
  • How do you think of a startup work experience when you are interviewing for Coursera? Skills used every day in a startup may be very different from skills learned from academia. Do you prefer a professional with years in startup or a new graduate from a Phd program?


Thank you for the excellent questions, Eve! My thoughts below:

  • I noticed that you have an interesting experience in Zipfian Academy's program. How do you think of this kind of bootcamp experience in your career?


In my case, attending a bootcamp was quite valuable. I had a lot of "theoretical knowledge" from college, but the bootcamp allowed me to transform that theoretical knowledge into "real world" experience. Additionally, it was a phenomenal way to build a network of aspiring Data Scientists. (This network is what helped me get a job.)

That said, I attended a bootcamp in 2014 -- back when the "Education Ecosystem" for Data Science was decidedly less developed. Now, there are many more online courses, part-time bootcamps, and Master's programs that probably satisfy the same needs I had.

As such, if I were embarking on my journey today, I'd probably weigh the bootcamp option against others, optimizing for the following criteria (in order):

  • Curriculum (with an eye toward what is most practical)
  • Network
  • Job opportunities
  • Do you think it is necessary to read lots of papers to become a real data scientist? If so, how did you teach yourself reading these papers and applying them to your job?


I think the ability to read technical papers is definitely quite helpful (although I think it's a skill anyone can learn)! In particular, when facing a "new" problem at work, chances are high that someone, somewhere, has written a paper on it. The ability to find that paper, read it, and implement it would certainly set you apart!

(At Coursera, we have a reading group that meets weekly to discuss a new paper that someone has either applied to their work or simply found interesting.)

All that said, I don't know if it's a "hard" requirement for all Data Science roles. I imagine it would depend on the specific problems you're solving.

  • How do you think of a startup work experience when you are interviewing for Coursera? Skills used every day in a startup may be very different from skills learned from academia. Do you prefer a professional with years in startup or a new graduate from a Phd program?


You're quite right in saying that the skills acquired in the two settings are quite different. As such, we usually have separate hiring pipelines for new grads and for people with industry experience 🙂
Userlevel 2
Badge
Usually a career break is a red flag for hiring managers sometimes is relevant to older workers, say 50+...
in your experience, or your perspective, what are the factors that would make you consider an applicant who had a career break of 5 years? What are the things that would make you DQ this candidate outright?
Likewise, what are the expectations for a career transitioner, whose prior experience is in engineering, but not in data science, nor directly in code forging?
Finally, what is the list of - maybe 5 - challenges that a hiring manager would need to overcome to put such a candidate into consideration against the continually employed, younger, coding adept, current, non-transitioner? Is there any value in a tangential perspective, coming from an unrelated but still technical/mathematically adept discipline (electrical engineering, or quality engineering, or accounting for example)
No sugar coating is necessary. Thank you.



I think that's an excellent question: thank you so much for asking it.

I mentioned this is another answer, but I think the single biggest challenge you'd face is getting screened out before you ever have a chance to speak with someone and prove your competence. As such, I would over-optimize for "How do I get past the resume screen?"

(The tactics I mentioned previously were (a) built out and lean on your network, and (b) create a portfolio (time permitting). However, there are undoubtedly others, whether it's doing a "stepping stone" job for a year or two, pursuing additional education, etc.)

Once you're past the screen, I think it's always easier to tell a compelling story. It's possible to put a career break into the proper context, and to explain why you're making the transition. Additionally, it gives you the chance to tell a compelling narrative about how your past experience (e.g., electrical engineering) makes you uniquely valuable for this role. (Note: There's always a way to spin why you're past experience is uniquely valuable for the role you're interviewing for 😉 )

A stray thought: Unfortunately, hiring managers are all too human, and bring with them a host of biases and misperceptions you'll need to overcome. For example, a hiring manager could look at an experienced "transitioner" and think, "Ugh, are they really going to be 'coachable', or are they stuck in their ways?" This is decidedly not a fair thing for them to think, but even so, it may make sense to address their concerns head-on. A potential story could be, "Hey, I know this is a new career for me, so I'm thinking of myself as a new grad, ready to learn." (Note: I'm not positive if this is the best angle, but something like it might work.)*

In any case, I hope this was helpful. Speaking as honestly as I can, I know I probably didn't do your question the justice it deserves. It's a critically important issue, and one that I think the technology industry should do a much better job of addressing.

*I should also add: I would defer to the career coaching resources that are out there on this topic, as I do not consider myself an expert here 🙂
Userlevel 2
Badge
Hi Dan,

Here's my question: In your opinion, how old is too old to switch over and get hired as a Data Scientist, not only in Coursera but in other companies as well? 35? 40? 45?

I'm asking because I attended a couple of webinars in the past where career switchers with significant IT experience are frustrated that they could not land a job in AI/ML field because, and in some cases they were told, that they are too old. I'm in the same boat, and want to learn more, but with this in mind, I cannot help but have second thoughts.

Thanks a lot, and have a nice day!

regards,
Alan


Hi Alan, thanks so much for the question!

I shared some thoughts on a related topic above, but in short, I don't think any age is too old. (Admittedly, I had this perspective instilled in me from a young age: My dad was an immigrant, and changed careers a number of times as he became more and more familiar with the job opportunities available here.)

Expanding a bit: One of the great benefits of working in a growing, in-demand field is that companies are hungry for talent, and so will be more likely to consider "non-traditional"* candidates. While it may require crafting a compelling narrative (see my answer above), I truly believe it's possible for anyone.

I've seen it happen too many times not to believe it :)

All that said, I don't want to trivialize the amount of work involved. Switching industries is hard. (When my dad went back to school to become a teacher, the only reason he could was because my mom could support us financially.) That said, assuming you're able to learn -- and there's more learning options available today than ever before -- you can absolutely make the switch: Many of Zipfian classmates were north of 30 and 40.

Thank you again for the question. I hope this was helpful!

*Honestly, given that Data Science is such a new field, I would question the very notion of "traditional" in the first place!
Userlevel 2
Badge
Hi Dan,

Thanks for your time and effort!

I have few questions how to merge data science and business:

  • How you translate the output of the data processor / model into business insights?
  • How seamless is this transition?
  • To what extent human aspect play the role (e.g. management does not understand math behind algorithms etc)?
  • How you overcome those challenges (trainings, use cases etc)?
Thanks a lot,
Best regards,
Misha


Hi Misha, thank you so much for the great question!

In my experience, this transition process requires quite a bit of work. After all, it's easy to produce graphs or try different model specifications, but distilling that into true business insight requires something much more difficult: an eye for what's really important.

Thus, when I'm presenting my work, I try to make it as simple as I possibly can. No jargon, no recapping of my analytical or model-building process. Just what I found, and why it's important -- stated as simply as possible.

Your mileage may vary, but in my case, I haven't actually found business or product stakeholders to get overly caught up on methodology. They generally trust that we know what we're doing. Rather, they notice when results seem unintuitive or surprising. Thus, I think it's the hallmark of a good Data Scientist to anticipate those moments and have back-up information. That way, when you're presenting, you can say, "Hey, I actually thought about that, but here's what I found, and why the result still holds..."

Over time, those little moments build a ton of trust.

One additional parting thought: To get an analysis to really stick with your stakeholders, I think you have to reiterate it again and again (and again!). As a rough rule of thumb, I'd say you should communicate your analysis three times more than you think you should 🙂
Userlevel 2
Badge
Whew, what a journey! Thank you again for all of the wonderful questions all. This was a ton of fun 🙂
Userlevel 7
Badge +13
Thank you for inviting questions and sharing your insights with the Coursera data science community, @Daniel Saber. We appreciate the time and thought you put into these answers!