Reading Time: 4 minutes

I collide with stories like these all the time:

“I had no idea how to write code two years ago. Now I’m an A.I. Engineer.”


“In April, I graduated from [some bootcamp] after 12 weeks of intensive data science training.”

Graduated? After only 12 weeks?

I feel conflicted when I see these. On the one hand, I’m elated to see our field growing with a much-needed influx of talent. On the other hand, however, I’m troubled by the number of junior data scientists and analysts applying for data scientist jobs with little experience.

Over the years, I’ve hired and worked with some extremely bright young data scientists who came from bootcamps or even transferred into data science from DevOps with very little or no experience working with real data. Real data always has warts and inconsistencies and myriad other problems. And it takes time to learn how to work with real data. Certainly longer than 12 weeks. It takes years. One of the wonderful parts of hiring a senior data scientist is that they need so much less oversite: read, they make fewer mistakes because they’ve seen more. They know what to do.

I should be clear, this lack of real-world experience is not any fault of the students or the bootcamps. It’s because there’s such a dearth of good, experienced data scientists while the demand for them is high, that the bar for what qualifies one to be a data scientist has dropped. But data science is hard. It requires the knowledge of far too many tools, and a solid grasp of math fundamentals.

I’ve seen applicants with two years of real experience applying for a senior data scientist position. This is unfortunate because occasionally someone with three or four years of experience (still not enough for a senior position in my book) will have to interview a candidate for a higher position, but with a lot less experience.

How would you feel if you were asked to interview a person who would likely be your boss, only to realize this candidate is your age, but with only two years of experience and you’ve got four? That’s a tough one, but I’ve seen it happen many times.

What can we do?
There are a few things we can do to ensure hiring qualified data scientists. For one, we can interview better. Project-based conversations are my preferred method of assessing whether or not an applicant will thrive. Furthermore, we can also let candidates know what we expect from them in interviews. Proficiency in data science, like many fields, comes down to understanding and applying concepts, not specific tools.

Tools Change, Fundamentals Don’t
Strong candidates will have the following attributes I always look for:

  1. At least one real hard-earned portfolio project 
    This would be an analytics project they did that demonstrates their passion for data science. I always prefer it if the data was a mess before the candidate got involved and I love hearing about what they did to remedy this. Then I look for a deep understanding of the project’s objective (the “why”) and finally I want to be sure the candidate can articulate the tricky parts well; if they can’t do this, how can they do it once they’re hired
  2. Real-world experience 
    Real-word experience outside of a bootcamp project on a few different projects. There’s an incredible amount of diversity in data science; I always want objectives.
  3. Genuine curiosity and self-sufficiency
    Genuine curiosity and an ability to “figure it out” on their own. We do so much of that, it’s almost comical. This is an essential skill for a data scientist.

Instead of posting job requisitions that list of 50 different technologies we expect data scientists to be proficient in, why not make it simpler and much more powerful? Require candidates to discuss a hard-earned project they care a ton about and ask them to clearly articulate their challenges and how they overcame them. These are great topics for a candidate at any level of data science.

What about the Fundamentals?
How do we teach fundamentals to the growing number of aspiring data scientists who didn’t receive years of torture, rather, university classes, that most of us had to endure? Mentorship and company-sponsored university classes can definitely help.

With genuine curiosity and a few years of real experience, I think most aspiring data scientists can go from student to data scientist, but a good grasp of fundamentals certainly doesn’t come from twelve months of training.

What do you think? I’m taking a pretty strong stance. Do you have a star data scientist in your team who broke the mold? Hit reply and let me know! I’d love to hear about them.

Of Interest

Who Doesn’t Enjoy a Good TED Talk?
Here’s a collection of TED talks for data science ranging from topics on the beauty of data visualizations to how your company’s data could end world hunger. Enjoy!

How can Algorithms be Sexist?
Algorithms have nearly mastered the human language. But why can’t they stop being sexist? To fight gender bias, researchers are training language-processing algorithms to envision a world where this bias doesn’t exist.

How A.I. can Ease the Pain of Booking Your Next Vacation: invests deeply in A.I. to help bring high levels of personalization to your experience using their site. Here’s an interview with Gillian Tans, Chairwoman for the Amsterdam-based, in which she discusses how the use of A.I. can ease the pain of booking your next vacation: