An Interview with Mike Tamir of Galvanize U: What it Takes to Become a Data Scientist

My name is Mike Tamir. I’m Chief Science Officer at Galvanize and Head of Education.

In order to be a top performing data scientist in the industry, you really need a number of skillsets to be mastered. We at Galvanize U bucket these into three major groups. There’s the theory – what is the machine learning, what is the mathematics, the statistics, the statistical modeling that you need in order to actually build these models and get the predictive algorithms that you need, and the data engineering – the scalable and distributed systems that really can create a high quality production product.

Number two, you need that practice. You need to actually get your fingers dirty, so to speak. You need to put them on the keyboard and actually code up these projects and get really good at implementing the things that you’re learning about when you’re doing just the theory.

And number three, you need the experience. You need to actually work with experienced data scientists, and you need to see them in action with implementing the craft.

You also need to practice doing that yourself. You need to get the experience as you develop those skills in order to become the kind of data scientist that every hiring manager is looking for. Every couple of months you will read about in the New York Times, in the Wall Street Journal, an article about “Why can’t anybody find these experienced data scientists, the kind of data scientists that people want?” Our philosophy here at Galvanize U is that nobody has gotten the recipe right. Nobody has figured out a master’s program that has the theory, but also has enough ample practice and enough of the experiential element to really get students coming out of the program to the point where they can be a top performing data scientist. The way we’ve designed this program is really focused on addressing these three core needs.

So it’s very common to want to know what it is you need in order to enter the program at Galvanize U. The strict requirements are that you must have a bachelor’s degree and because it’s an accredited master’s program, you must have that bachelor’s degree in order to be eligible for the master’s. What the degree needs to be in is a STEM program, so science, technology, mathematics, engineering, but broadly construed.

So an economics degree, a finance degree, these sorts of degrees are actually very, very helpful and a lot of data scientists actually come out of programs like this. So we’re actually going to be casting a wide net when evaluating what kind of training you’ve already had. As far as what skills you might want to either already have, or you might want to develop and learn a little bit about in anticipation for going through the application process, linear algebra and statistics is fantastic. It’s really the lifeblood of the mathematical understanding of a data scientist.

So refreshing your skills on that or maybe checking out some of the online resources to help, you know, get orientated in these subjects could be very helpful. And the program will take you from having taken it in the past, all the way to being an expert and there’s an entire component of the curriculum focused on developing these skills at the advanced level that you need them to be.

A common question is, “How long does it take to become a data scientist, from the bachelor’s level all the way to a high performing professional expert?” I’ve interviewed and hired dozens of data scientists in the Bay area over the last several years, and one thing that is a very common pattern is if you’re coming out of a traditional theory focused – or I should say, an exclusively theory focused data science program or machine learning program – there’s a ramp-up time, by the time you come out of that program to when you’re ready to be fully productive as a top data scientist. Often it takes as much as six months of on-the-job learning before a straight out of grad school student is able to perform at the levels that we would expect them to be performing at.

So what sort of skills, in particular, should a student focus on? The answer I usually tell an upcoming student is that you need to have a little bit of everything. Data scientists, even a data scientist going in to be a pure theoretical modeler for an R&D company is going to need to have the conceptual understanding of the data engineering and the processing skills that are going to come into play when the model actually goes from prototype to production.

On the other side, even a data scientist who is focusing more on the data engineering and scalable implementation sites needs to understand the machine learning and the mathematics behind it because in order to implement that successfully, you need to understand what’s going on. And so you can help to make sure that the results that were gotten in prototype stage are actually preserved and maybe even improved by the time they get to production.

And last, you need to be able to communicate. It’s a common, common issue that somebody who has deep understanding maybe just in the machine learning or just in the data engineering, can’t necessarily communicate that understanding or the effectiveness of the model well, and that can lead to infelicities and trying to actually get your high power, really great product, or really great prototype to the floor and actually see it implemented. So being able to get the core skills you need for data communication is also very, very important for the most successful data scientists.

It can be very daunting when you’re trying to break into any new field, in particular, a technical new field, and data science is no different. The one piece of advice that I would give any aspiring data scientist is to stick with it and keep going. Don’t lose heart. It’s very common, in fact, it’s usually a good sign, often a student will hit a point where they realize how much they don’t know yet, or how much there is to learn. And there’s a moment of almost panic where, “How am I ever going to get around this?” So my advice would be stick with it, keep going, but also start leveraging those resources that are out there.

There are a lot of great online resources. There’s a lot of great data science communities out there. There are a lot of great data science hubs, like the one here in Galvanize, San Francisco, where there’s tons of people who are willing to help you get through, get orientated, go over the hurdles that you might be daunting you right now. And then you have the exciting end of the tunnel to look forward to where you say, “Wow, I’ve really accomplished something. I’ve really mastered this field,” and it can be very satisfying.