Computers aren’t what they used to be in the 1980s. Take it from Dave Joffe, the chief data scientist at Bank of America’s control function technology department. He’s seen how technology, data, and computer science have developed over the decades—and ultimately converged.
Joffe has been a data scientist for most of his career, though his roots are in a different type of science. He graduated with an undergraduate degree in psychology before discovering a fascination with computers while working in a neurobiology lab. Joffe first learned to program on the lab’s microcomputer, “which started a whole path I couldn’t have foreseen,” he tells Fortune.
After working on the microcomputer in the lab, Joffe decided he wanted more formalized training and went back to earn his master’s degree in computer science. That landed him his job as a programmer with Bank of America—where he’s steadily grown into a data science leader over the years.
Learn more: Is an MSDS worth it?
In 2020, there were an estimated 2.7 million open data analysis or data science jobs and a 39% growth in employer demand for data scientists and engineers, according to IBM. While demand for these roles is high, top companies like Bank of America still look for high-quality candidates to fill open roles. In fact, Joffe recalled a year-long search for a candidate to fill a data science role at Bank of America.
What can aspiring data scientists do to prepare for the ever-changing world of data? Fortune recently sat down with Joffe to discuss important considerations for future data scientists.
The following interview has been edited for brevity and clarity.
Fortune: What were the options available to you when you went back for your master’s degree, and why did you choose what you did? How does that differ from today?
Joffe: Yeah, very different. I just knew I needed a more formal training in what I was doing, which was programming. And computer science seemed to be the closest to what was happening then, but that put me on a very particular track and I wouldn’t have even thought about the other track that was present at the time—but now they’re merging.
I could’ve gone into a mathematics, applied mathematics, or statistics program—I could’ve gotten a master’s degree in math or statistics. At the time, there was no way for me to know that was highly related to the programming that I was doing. I just couldn’t have known that.
Fortune: What is your advice for data science professionals today? Should they pursue a master’s degree?
Joffe: Notwithstanding the amount of cost that all these things are—that’s a whole separate discussion—they were much less expensive when I did that. Nonetheless, the answer is yes. That knowledge is important. So is experience, and so are skills. But they’re all hybrid.
An advanced degree is very helpful in building a foundation of knowledge. But it’s going to depend which master’s one does and depends on which of the hybrid—roughly four areas, domains, or disciplines—you’re interested in.
One can have a master’s in computer science is still applicable, especially in the area of database programming because the nature of data science is that it’s data hungry. So you better be a very good programmer with very large amounts data. Data science is applicable to all of them, but computer science would be very strong programming skills.
Second is the whole area that I mentioned before: statistics and machine learning. That would be more, say, a master’s in mathematics, applied mathematics, physics, or some are even in biology.
Third, what about business domain knowledge? Data science is an interesting term. It’s a term with which I actually disagree in that it really is science of something else of which data allows them to do the science. It’s not the data itself, the science of data information theory that’s in question.
In the case of Bank of America, it’s the science of banking, which falls under economics and finance and the application of the programming and the deterministic and probabilistic programming skills. The first two make this useful. So if you’re about business domain knowledge and you want to build a foundation there, an MBA with business analytics, a master’s in business analytics all seem appropriate for that domain.
Fourth—this is the most scarce of the domains—we don’t find many people hybrid-skilled here. That’s that these are such complex fields that the ability to communicate and visualize results to match these complex artifacts is probably best pursued by the hybrid of all of them, which is a master’s in data science or an MBA with analytics.
Fortune: What does Bank of America look for in top data science candidates?
Joffe: Certainly we do look for those degrees, but I’d say that any undergraduate with a STEM background of various sorts coupled with machine learning experience—machine learning I would estimate is greater than 90% of all of the techniques used in data science. There’s a revolution happening.
If someone has a lot of machine learning experience … we like that, too. We already know that we need to accept folks who are going to be weaker in either deep knowledge through advanced degrees, and we internally train a lot.
Fortune: What other continuing education opportunities should data scientists pursue?
Joffe: I think that, in general, if you’re a data scientist at Bank of America, Bank of America has extremely valuable data and we value that data. And so they can gain that experience. So it’s not just the training, it’s on-the-job. What I think is extremely important for anyone beyond certification and degrees is gaining skills, knowledge, and experience and finding valuable patterns in data. This is key.
That’s how you grow in value. It turns out that those who can find valuable patterns in data become valuable themselves. That means, though, that they want to be at a firm that has valuable data, which is value-laden so that they can gain that experience.
Fortune: What other advice do you have for aspiring data scientists?
Joffe: Lifelong learning is important to anyone who actually wants to go into this field because the nature of these fast-growing areas is that they evolve very quickly.
So lifelong learning in the areas that we’re in right now, which include all of the great work that’s going on in language processing, or unsupervised learning, or reinforcement learning, or GANS (which stands for generative adversarial networks), or complex systems, which have become very much in favor with climate modeling right now. Fairness testing, explainability—which goes along with the whole ESG company’s focus on environment, social, and governance issues.
All of these are evolving very quickly. You don’t have to know all of them, but one will have to learn to keep pace with this evolution continuously.
