Want a Job in Data Science? Here’s Why You Should Specialize
If you’re reading this, chances are you’re looking for “a data science job. ” But what does that actually mean? The data science field consists of more than just data scientists. A company might be hiring for an analyst, an algorithms developer, a machine learning engineer or a data engineer, among other roles. Which are you?
“I can do it all!” is a common answer, but it may not be the best one. Being a jack-of-all-trades might be necessary if you’re joining a very small team where you’ll be the only “data person” on staff. But when you’re looking for data science work, it usually helps to have a specialty.
Why Specialists Get Hired
When it comes to working in data science, few people have as much experience as Mike Kim. He’s the co-founder and CTO of Outlier.ai, and he’s led data science and machine learning initiatives at Google, Aardvark, and AltSchool. As someone actively hiring for roles in the data science field, Kim said he typically looks for applicants with very specific skills, not data science generalists.
“This is what generally just makes data science hiring, and job seeking, fundamentally hard. It’s that the term data scientist encompasses so many different kinds of skills, and kinds of people,” Mike said. “It depends entirely on what it is that I'm looking for.”
For example, let’s take a look one of his most recent hires: a machine learning engineer. While the job description certainly falls into the realm of data science, hiring a data analyst or even a data science generalist doesn’t make much sense if it’s possible to find a candidate with the specific machine learning skills Mike needs.
Those specific skills color how Mike is going to assess applicant resumes. “I need to see the buzzwords around large-scale machine learning platforms, or whatever these things are. If all you list are things like, ‘Oh, yeah. I write SQL and I write reports and I can build dashboards in Tableau,’ you're not the person I'm looking for.”
The reverse is also true, he said. “If all I am really looking for is someone to be an analyst, then I don't want the person who's built a massive Spark cluster, and run it on these specialized models.”
Kim’s approach isn’t unique. Hiring managers all have to navigate how to find the right candidate for the right role in an industry with lots of technical overlap and a ready supply of generalists, particularly at the entry level.
“Just imagine that you’re a company trying to hire a data scientist,” Harris writes. “You almost certainly have a fairly well-defined problem in mind that you need help with, and that problem is going to require some fairly specific technical know-how and subject matter expertise.” So why hire a generalist if a specialist is available?
On the other hand, it’s worth keeping in mind that companies don’t always know that they need a specialist, or know what kind of specialist they need. This is a pretty common problem in a field as new and fast-growing as data science. One of the challenges of applying for data science jobs is assessing whether the company actually does need what it says it needs, and to what extent your skills can actually help solve its problems.
If you’re a machine learning engineer, you don’t want to take a position as a “machine learning engineer” who ends up spending all day doing basic analysis because the company didn’t actually need (or doesn’t have the data to support) machine learning. Unfortunately, there’s no easy solution for this; it’s something candidates will have to try to assess for themselves during the application process.
Emphasizing Your Speciality
If you’ve reached the point of applying for jobs in data science, you’re probably already aware of some of your own strengths and weaknesses, as well as the sorts of projects you like doing. This can suggest a good direction in which to specialize.
But even when you’re aiming to be a specialist, you’ve probably got a quiver of general data science skills, and it can be tempting to try to sell yourself as a jack-of-all-trades just in case that company looking for a machine learning engineer might value someone who can build dashboards, too.
Fight that impulse. Take a detailed look at the job description you’re interested in, and tailor your resume and portfolio to provide examples of your skills and projects directly pertaining to the role. A hiring manager might be looking through thousands of applicant profiles — your relevant skills should be clear at first glance, and not surrounded with irrelevant work that makes it less clear who you are and what you do.
“If all of your portfolio is ‘Look at these cool dashboards I've built,’ then I know that you're one kind of data scientist,” Mike said. “If all of your portfolio projects are ‘I built this really gnarly classifier, or this data set,’ then I know you're a different kind of data scientist.”
To be clear, there’s no wrong sort of data scientist to be. But the dashboard-builder wouldn’t be a good fit for a job designing machine learning models, and vice versa. Moreover, if an employer can’t quickly tell that you’re the specific kind of data scientist that they need by glancing at your resume, chances are you’re not going to get a call back.
What Else Can Set You Apart?
Once it’s clear you’ve got the specific technical skills required for the position, Mike said, he moves on to assessing other qualities.
One important one is grit. “A lot of data science is grungy data gathering and cleaning, testing and validating -- much less glamorous or fun or exciting work, but incredibly important,” Mike said. “I definitely look for candidates who have an understanding of this reality, and better yet, firsthand experience or evidence of being able to handle this well, with the right attitude about it.”
Another one? “I prefer candidates from rigorous quantitative backgrounds. I don't care what field that was in — but I prefer people from quantitative backgrounds,” he said. Failing that, he said, he wants to see “something in their experience that has shown me that they have the quantitative skills in practice.”
This doesn’t mean that you can’t do data science with a background in the humanities (we’ve seen that that is very possible). But it does mean that if you don’t have a formal quantitative background, you’ve got to do everything in your power to prove that you have those skills and can put them into practice, even though you didn’t go to school for them.
Mike also looks for applicants who have communication skills that go beyond just collaborating with technical and nontechnical people. He’s looking for a blend of tenacity and flexibility.
“The odds that you come in, and have everything I need you to have are basically zero, right?” Mike said. “So the ability to be flexible, be practical, take feedback, and to adapt to it — and to grow...that is really the key thing I look for.”