As a student interested in data science, it’s not always straightforward to know exactly what you should focus on to get that first data scientist position. Also, being one of the faster-changing career paths, it’s not always clear when certain pieces of career advice have become a bit dated or are even applicable in 2017.

I reached out to several data science leaders of various backgrounds to get their thoughts on this; below are their responses to this question:¹

What advice would you give to students that are interested in becoming data scientists?

Simon Petit, Cofounder at dataroots:

“To my opinion, students willing to become a data scientist or to work in this specific area should be naturally curious in life, eager to learn new things and not scared of thinking out of the box.”

“Secondly, they must develop a strong background in applied statistics and mathematics combined with programming skills to be at ease in the different techniques of analytics later on.”

“Third, they need to go beyond their expertise and be able to understand and adapt to any business they work for. Last element, communication and social skills are very important when talking to clients (business owners) and explaining the models and their benefits.”

 

Dan Valente, Head of Data Science at Knotch:

“Be curious, be skeptical, never stop asking questions (even when you think you have the answer!), learn as much math as you can, and get very comfortable programming.”

 

Note: Dr. Saigal is giving advice that is specific to high school students.¹

Dr. Sanjay Saigal, Executive Director, Master of Science in Business Analytics at UC Davis:

“I tend to find that high school students don’t suffer from a lack of technical education as much as a lack of curiosity. That is to say, being creatures of culture (like all the rest of us) they don’t very often seek to cultivate the scientific temperament.”

“Were I advising high schoolers, I’d recommend that they look for opportunities to investigate truth – whether be it using methods of analytics or chemistry or any other science. Of course, learning statistics and computing helps discover truth too!”

 

Emily Glassberg Sands, Director of Data Science at Coursera:

“The technical skills are just one piece of the puzzle – necessary to be good but not sufficient to be great. Push beyond the technical. Think deeply about the product and business sides of the challenges you’re trying to solve.”

“Find a company where you care deeply about the end goal – the product fascinates you, the social mission speaks to you, whatever. You deserve to wake up every day excited about both the how and the why.”

 

Anahita Hassanzadeh, Data Science Manager at The Climate Corporation:

“Get your hands dirty with open-source data challenge questions such as the ones on Kaggle. This will help you gauge your strengths and weaknesses and also will make your resume stronger.”

 

Bill Vorhies, Editorial Director at DataScienceCentral:

“If you’re interested in becoming a data scientist you should look specifically for a college that has a data science curriculum, not just computer science.”

“Think also about taking business courses or getting specific business experience in the industry you’d be most interested in since data science is about solving business problems and creating business value, not about math or computers.”

“Mastery of predictive analytics will get you 8 out of 10 data science jobs but if you want to work at the cutting edge, prepare for a Masters that focuses on deep learning using neuromorphic or quantum techniques. Both those will be coming strongly on line over the next four years and will be in great demand.”

 

In Closing

Thanks to all our contributors for sharing their thoughts!

 

 

1. Note: The contributors quoted for this article were asked slightly different subtypes of this question (ex: advice targeted to high school vs college students).  After going through the responses, I thought it’d make more sense to combine the responses and put them all in one article.  Any errors and omissions are my own.

 

I get it – you love digging into the data. The scientist in you loves being on the frontier of knowledge, and you’re always (at least feeling) like you’re on the verge of a breakthrough insight. The whole process of getting your hands on the raw data, the (non-always-so-fun) process of getting that data into a somewhat usable form, making your first hypotheses…the list goes on and on.

And let’s face it; this usually translates into long hours of high-intensity focus, where proper rest and recovery take a back seat (at best).

 

Pushed a Bit Too Hard

It’s a great feeling while you’re making progress, starting to see the underlying structure of the problem take shape; getting glimpses of what the final product and insights might look like. And there’s nothing wrong with putting in a substantial effort – after all, that’s what you’re paid for.

However, there is a limit – not just for what is reasonable for your boss/client to expect from you, but what you can expect from yourself.

For those of us who have not yet experienced professional burnout, it’s hard to cleanly explain what exactly it is or why it’s so bad. However, it’s becoming a major issue for data scientists: One study showed that over 25% of data scientists say they are ‘heavily stressed’

More than 25% of data scientists say they are heavily stressed

Basically, at least for me, it’s when you’ve pushed yourself too hard over a months-long process and you’ve noticed your motivation fall off a cliff. Your brain seems to have lost 20 IQ points and you can’t quite get yourself to care anymore about what you started in the first place.

You start feeling like you’re stuck in a fog, and the mini-breakthroughs that usually come so naturally (and frequently) start to become further and further in between. In other words, it’s a not fun position to be in, and it’s not always a quick fix to get out of it.

 

Prevention Focus

In my opinion, the old quote ‘an ounce of prevention is worth a pound of cure’ really comes into play here. Preventing burnout could be something as simple as minor tweaks to your work routine – while recovering from full-on burnout can take months, or even kick off a quarter- or mid-life crisis.

One way to think about burnout is like a stress fracture – repeated stress with insufficient time to allow the (individually benign) microfractures to properly heal up. We all like to think of ourselves as mental superhumans, but we each have a limit to how far we can push ourselves before our brain essentially shuts it down and demands time to ‘heal.’

For data scientists in particular, that implicit belief about being mentally stronger than average is a major risk factor for burnout – we think we’re invincible. That passion of finding the hidden relationships within the data, it’s so hard to turn off.

We are so busy solving awesome problems that we never really take time to consider whether we’re pushing ourselves too hard. And often, we don’t realize it until it’s too late – as in, a weekend trip isn’t going to be enough this time to recharge our batteries.

So what can we do to actually prevent and/or recover from burnout? Although I’m nowhere near an expert on the topic, from my (unfortunate) experience I have some thoughts on some things that at least worked for me.

Depending on who you ask, the personality traits of top data scientists can apparently span a huge range.  However, when looking at these opinions in aggregate, some commonalities start to emerge in terms of what really matters and separates the average from great data scientists.  From my perspective, here are the three key categories:

 

Curiosity

In my opinion, this is both the most important factor while simultaneously being the hardest factor to learn.  At their core, what data scientists want to do is take lot of data and find the relationships that no one else has yet found.  

This is a lot where the term ‘data scientist’ originated from – scientists fundamentally try to be on the active frontier of knowledge and find out things that haven’t yet been discovered. Without this innate desire and trait, it’s hard for someone to make a career as a scientist, let alone as a data scientist.  When someone talks about a data scientist having ‘passion,’ this is what they’re talking about.

We are trying to prove ourselves wrong as quickly as possible, because only in that way can we find progress. -Richard Feynman

One signal to roughly estimate someone’s innate curiosity is to discuss what they do in their free time.  Although there are certainly many exceptions, it could generally be considered a promising sign when their hobbies involve something science-related (that is unrelated to their day-to-day job).  

A subset that I believe falls within this category is statistical thinking.  Essentially, knowing whether you’ve made a true discovery is intimately linked to understanding the significance of your finding – especially when you’re dealing with messy data, as most of us are.  

The good news is that there are solid training options available for those wishing to concretely improve the quality of their statistical mindset.

 

Communicator

Unless they don’t have to convince anyone of the value of their work, communication is also a critical skill for a data scientist.

It’s especially tricky for data scientists to be good presenters, as they have to simultaneously be the expert in the technical minutiae of their project, while also be able to quickly and efficiently explain what their findings actually mean to a potentially very non-technical audience.

This dichotomy of required skills (great technically, great communicator) is one of the driving factors behind why data scientists are one of the more highly-paid professions.  It’s relatively easy to find someone who is either great at digging into the data or great at presenting findings, but finding someone who is good at both remains a key factor in the huge, unmet demand for top data scientists.

Now that we have huge machines to absorb huge amounts of data, the value of big data is clearer. -Andrew Ng

Fortunately, unlike curiosity, it’s much more common for someone to be able to improve their technical communication skills.  Results will vary, but there at least exist concrete frameworks to become a better presenter.

 

Technical Acumen

I use this term broadly here, and I rank this trait as below the first two for this reason: it’s the most straightforward to learn – at least to a passable degree.  

Whether it’s the company offering the specific technology framework providing training or a colleague giving a hands-on crash course, learning how to be at least moderately effective in a new technology framework generally one of the less-rare skills amongst data scientist.  

Of course, what falls under ‘technical acumen’ and the levels of talent span huge ranges, and all else equal the more technologically proficient someone is, certainly the better.  The most straightforward way to improve this trait is what we’d expect – write code everyday, experiment, fail, fix, and repeat.  

Be on the lookout for tips and shortcuts from the relevant community, and don’t be shy about at least trying their suggestions in your own work.

Most of us have either conducted an interview for a prospective data scientist hire or been interviewed for a data scientist position.  A lot of the focus on data science interviews focuses on the common technical questions, and rightly so.  

However, sometimes the key factors for the potential fit are solidified by answers to the non-technical questions, which themselves can quickly evolve into highly technical answers. Some of the more useful non-technical interview questions are listed below:

 

What is the fundamental role of a data scientist?  Can you give an example?

The main reason I love this question (or question set) is it’s a great indicator of whether someone can explain highly technical work to a very non-technical audience, except in kind of reverse-format.  Instead of taking something technical and framing it in general terms in order to be comprehensible to the audience, this time they’re starting with something very general and gradually working up to some potentially highly technical concepts.  

Their domain experience and communications skills are just as important as their technical skills. -Wayne Thompson

Starting from conveying a high-level understanding of why data scientists are valuable and then distilling that down to a concrete yet still decipherable example, all on the fly, is a skill that many employers would say is lacking in most of the candidates they evaluate.

Additionally, by asking the second part of the question in asking for an example, you get a chance to get an early impression of whether the candidate is very strong on marketing but lacking on substantive and concrete experience.

 

If you join, what’s the first question you’d probably ask our head of data science?

This question is a bit more off the beaten path, but again could provide some unique perspective on what the candidate is all about.  If they reply with some canned response like ‘what does your average day look like,’ that’s not necessarily bad.  

However, sometimes a candidate will have a very perceptive question that demonstrates far more knowledge of the relevant domain than you had thought, and this knowledge might not have been brought to light during the standard data scientist interview.

 

Which data scientists do you admire most?  

This is an interesting question and could go many ways.  For example, the candidate might be a top data scientist but then not really keep up on the individual news of what other data scientists are doing.  

They could be caught a bit off guard with this question, and how they react when they don’t know an answer that they think they should know is very telling.  Seeing how someone acts in a moderate pressure situation and thinking on the fly is a highly useful data point in this age of canned interview questions and responses (not to mention interview coaches).

On the other hand, if they start naming off some of the more conventional names (with no one that you haven’t heard of), you might ask a follow-up question about what specific aspects of the work do they admire?  

This is where you can potentially start to get an idea of whether the candidate has a tendency to confidently discuss items that they really aren’t at all familiar with.  

The data scientist who is openly willing to say ‘I don’t know’ is unfortunately becoming a more rare commodity, with a more common reaction being to start using big words to hopefully draw attention away from their lack of knowledge in a specific area.

 

We’ve all heard it – it’s nearly impossible to pry away top data scientists to join your team. It’s an exaggeration, but it’s generally true that top data scientists are already in a very comfortable position with a team that’s doing very interesting things.  How can a company, especially one that might be smaller and with fewer resources than the tech giants, realistically compete?

The market of data science jobs is, unfortunately, still dominated by buzzwords. -Dima Korolev

First off, the situation isn’t necessarily that dire.  Assuming you’re willing to pay at least market rate, data scientists are generally most interested on the type of work they’d be doing, versus the brand name of the company.  

 

Not an Impossible Task

As a smaller company, you’re likely already in a high-growth potential area, maybe riding on a trend that could be huge within the next few years.  This is this the stuff that data scientists really care about – as long as it’s conveyed properly.

In terms of describing the role to a prospective hire, don’t just focus on the growth of the company or team.  Sure that’s impressive, but that’s also there in probably 90% of the pitches we hear from recruiters.  

What’s much more important is making it very clear exactly what my role could be in helping with that growth – how exactly can I expect to contribute and grow with the company?  What will my day-to-day look like, and what type of direct career progression can I expect as the company grows?  

In what specific area will I be helping with, and how exactly is that linked to the growth of the company?  

 

What Keeps Them Passionate

Additionally, one of the core qualities of a top data scientist is curiosity – what hidden relationships will they find in the data that could have huge real-world impact?  After a few years at a company, some data scientists will start to feel that they’re getting a bit of diminishing returns in their current role.  

Financially they’re probably doing well, but intellectually they start to get a growing sense that something is lacking, and things aren’t quite as exciting as they used to be.

For data scientists, they are especially sensitive to that sense of excitement.  It’s what drives pretty much all of us, and when that passion starts to fade, it’s a huge driving factor for looking around for other opportunities.  Especially when we’re given the opportunity to get our hands dirty with fresh data sets that could be at the early stages of a huge growth industry – those are some factors which less-resourced companies should emphasize.  

 

Sources of Growth

Again, this ties back to what specific role the data scientist would play in that growth. Mission is great and all, but unless we see a very clear link between the potential day-to-day and how that directly impacts the company’s growth, it’s hard to get the attention of top data scientists.  

They already have all the stuff that you’re probably offering, and more – with the exception of getting in early as a key contributor/pioneer in a field that isn’t yet well-developed.  

Any recruiting and outreach process for data scientists should start with a brutally honest assessment of how exactly your unfilled role compares with the rest of the market – especially from an intellectual perspective.  

If you really feel stuck, reach out to data scientists you know (in a non-recruiting capacity) and discuss the job with them, and most of they will bluntly tell you what’s lacking in your pitch, and simple tweaks to make it more effective – and if you have no data scientist friends, feel free to reach out to me.  

In other words, the iterative process of getting (and being receptive to) real-world feedback could be more valuable than some hiring managers assume.

One of the more common approaches that a freshly-minted data scientists takes when given a project is to immediately fire up their programming framework of choice and apply one of the fancier algorithms they know.  

However, this initial tendency starts to diminish as you take a look at the habits of more experienced data scientists – why is that?

 

A More Big-Picture Approach

First, there’s nothing wrong with experimenting with a wide range of approaches when you are first exposed to a new data set.  After all, there isn’t necessarily one true method of getting your bearings when figuring out what exactly the data is saying, at least at a high level.  

However, some would argue that applying high-buzzword-content algorithms at the immediate outset of a project isn’t the best way to go.  Sure it seems like you’re making progress and sometimes you get really cool graphs – but how do you know whether that graph actually means anything?

 

Asking Initial Questions

This is where the distinction between rookie vs experienced data scientists starts to come into focus.  A more experienced data scientist will take a more circumspect approach when first digging into a problem – for example, what do we know about the precision and accuracy of the data?  What’s our general comfort with the cleanliness of the data?  Are some variables more prone to error than others, and within what order of magnitude is this error?  

Asking questions like these (and perhaps slowing down a bit) can have a major impact on avoid early local minima – or even worse, overfitting a model to an artifact of the data that wouldn’t have been that hard to avoid.

 

All About The Presentation

Additionally, something that a lot of newer data scientists might not quite adequately appreciate yet is the disproportionate importance of giving a quality presentation for their findings, and tying the findings to real-world business objectives.  

In other words, it’s great if you analyzed many terabytes of data with a clever deep learning algorithm, but no one really cares about that – it’s much more important if you can convey what your results actually mean.  

Your data may hold tremendous amounts of potential value, but not an ounce of value can be created unless insights are uncovered and translated into actions or business outcomes. -Brent Dykes

It’s easy to get so caught up in the data wrangling process that you forget to leave enough time to put together your analysis presentation – however, this is much less likely to happen amongst the more experienced data scientists, who understand that perception is often more important than reality.

Finally, more experienced data scientists realize that more and/or better data will beat fancy algorithms almost every time.

Often times, the initial data set we’re using to work with doesn’t represent the entirety of data that is reasonably available to us, and by directing maybe a few thoughtful questions to the data source provider, it might not be that difficult to immediately eliminate potential data cleanliness issues or errors.  

 

Limitations of the Data

Asking careful questions about known data issues is a trait common amongst more experienced data scientists, but for the newer data scientists, they’re generally more gung-ho and just want to start digging into the data ASAP.  

That enthusiasm is great, but without a basic understanding of the qualities of the data set, many a disappointed data scientist have frustratingly found themselves back at square one.

Practically, this generally means that the more experienced people will start out with a pretty basic visualization routine, just to get a feel for what they’re dealing with.  Simple graphs of the data, even in Excel, could quickly lead to the discovery of high-sensitivity data issues that the data provider wasn’t even aware of, but could quickly fix or be adjusted for.  

A lot has been written lately about which type of data science and analytics program makes the most sense for your situation.  However, the press has been a little more sparse about exactly what the post-graduation job hunt looks like – especially in that crucial first few months.

 

Early Action

Why are the first few months so critical?  Fair or not, potential employers are at least implicitly looking for some ‘social proof’ around their candidates – or in the case of a prolonged job search, the lack of substantial interest from any other company could send a cautionary signal.  

For better or worse, there is a premium associated with soon-to-be or fresh graduates (including traditional computer science university graduates), and that premium erodes over time.

 

A Numbers Game

Especially in a quickly-evolving field like data science (some would argue it’s THE most rapidly changing ), in extreme cases the skills and frameworks that were impressively relevant a year ago might not as much in demand after the initial buzz has died down a bit.  

So what does this mean for your job search?  As with most prospecting efforts, in large part it does come down to a numbers game – how effectively can you get yourself out there, especially in the first few months after graduation?

For me, getting a job in data science was 80 percent networking, 20 percent skills, and not the other way around. –Will Stanton

First, I wouldn’t necessarily recommend targeting Fortune 500 companies – companies with deeply ingrained HR departments that are looking for reasons to programmatically eliminate candidates from the hundreds/thousands of resumes they receive per day.  

Low-hanging fruit for them would be to automatically eliminate candidates whose most recent experience is in a data science bootcamp.  

 

Connecting With the Right People

Rather, the smaller and more early-stage companies are where most recent boot camp grads have been reporting the most success recently.  Essentially, these companies are much more directly interested in what skills you bring to the table right now, as opposed to larger companies who might be more interested in how smart they can make their hiring processes look to the board of directors.  

Additionally, a big benefit of looking for non-huge companies is it’s much easier to directly get through to someone with direct hiring power.  Instead of throwing resumes into a black hole for the larger corporations, you can expect responses from CEO and CTO-type people – and sometimes they will even interview you themselves – and sometimes very quickly.

 

Cutting Out the Middleman

When you’re directly interacting with the high-level people within the smaller companies, you also cut out the not-always-optimally-competent middleman.  

To get hired at a large corporation, you generally need to impress at least (a) the automated filtering system, (b) a HR generalist, then (c) the hiring manager.  When talking directly to people in smaller companies, steps (a) and (b) become much less relevant.

Shifting gears a bit, when it comes to the actual interview, there is still a good deal of preparation that is well worth your effort.  Especially amongst smaller companies there is a huge range of technologies/frameworks/domains that you’d have to become at least passably familiar with to have a solid chance of landing an offer.  

Fortunately, a lot of the founders are very open about exactly what they’re trying to accomplish and how they’re going about it – sometimes as granular as exactly what technology frameworks they’re using and what they’re trending towards.  

As with pretty much every company, interview prep is critical – the good news for the smaller companies is that it’s generally much more straightforward to figure out how to actually prepare.

After all the time you’ve spent diving into the data and progressing towards finding out what’s actually going on, it’s now time for you make a huge mental shift and figure out how you can actually present your findings and conclusions.  

Not only that, but you usually have a very finite amount of time to present, and the audience will mostly likely be non-technical.  

It’s a high-stakes position to be in, regardless of whether it’s your 1st time or 100th.  And as a lot of us have seen, being a great data scientist doesn’t always translate into automatically being an effective presenter.  

If all you want to do is create a file of facts and figures, then cancel the meeting and send in a report. -Seth Godin

The single biggest thing I’ve learned from those situations basically boils down to you have less presenting time than you think – especially if you’re taking questions.  

Generally, no matter how often you’ve practiced and gone through your presentation notes, you probably can count on at least one significant aspect of your presentation being far more technical than you thought – at least from your audience’s perspective.  

And for that one item, it usually ends up being very important for your audience to have at least some semblance of comfort with what you’re talking about, and this usually results in some back and forth discussion.

 

How Helpful (really) Are Your Graphs?

Not only that, but also as you’re presenting you might come to realize that one of your graphs is nowhere near as explanatory as you had initially thought.  You realize on the fly that one of the axes, which makes perfect sense to you, only makes perfect sense because you’re the one who freshly derived the y-axis proprietary variable less than a week ago.  

Maybe to your immediate manager your graph and axes make fluent sense, but to anyone not intimately involved with your line of thought (of why you had to derive that axis in the first place) will now be looking at a graph with little to no actual informational value.

All this to say, almost regardless of how big-picture you think you’re including in your presentation, it’s probably not enough.  At least for me, I’ve found that presentations have gone much more smoothly when I start with the almost basic high-level concepts that I know we’re all familiar with, quickly going through those, and then establishing a very simple link between those concepts and each graph that I’m going to show.  

I’m sure many of us have sat through at least a few painful presentations where the presenter makes a couple high-level remarks about their topic and then starts showing chart after chart after chart – but at most 5% of the audience actually understands what the presenter is talking about.  

 

Speaking Their Language

The sad part is that, with maybe two more minutes of highly-targeted remarks of linking the graphs to concepts the whole audience is familiar with, that 5% number would probably be much more closer to 90%.

At the C-suite level, they don’t really care that you have some fancy graphs and highly technical words.  What they do care about is revenue growth, churn, yield, client retention – the list goes on.  

Just taking the extra minute or two to make sure the audience is able to make the fundamental connection between (a) what they care about and (b) your graphs –  it’s something that might end up having the highest per-minute ROI of your entire data scientist career.