Dr. Scott Aefsky has a bachelor’s degree in physics and mathematics from Clarkson University and Ph.D. in physics from Brandeis University. Since completing his doctorate, Scott has worked in data science and software development at the private fusion energy company Tri Alpha Energy, defense contractor Frontier Technology, and Agero, a company that handles roadside assistance for businesses in the automotive and insurance industries. Scott recently took some time to speak with me about his career experiences as an industrial data scientist and share his advice to current graduate students:
Matt: Hi, Scott. Thanks for taking the time to talk to me today. I appreciate it.
Scott: No problem.
Matt: All right, these days you’re a staff scientist at Frontier Technology which I understand is a small defense contractor. What’s it like working there?
Scott: I really should’ve updated my LinkedIn page before talking to you. I actually don’t work there anymore, I left there a month ago.
Matt: So, where are you at now?
Scott: Now I’m working at a company called Agero working as a data scientist.
Matt: I think that, at the last couple of jobs you were working at, “data scientist” may not have been your actual title, but it accurately describes what you were doing, right?
Scott: Yeah, that’s right.
Matt: What is being a new data scientist at Agero like?
Scott: It’s good. I’m like two months into it so I’m still kind of finding my footing. What the company does for their main business is operate roadside assistance for insurance companies or car dealers that offer it. They’re kind of like the middle man. It’s a big company. They’ve got a lot of data. They’ve only recently, like within the past two years, started a data science group. There’s two aspects to the data science group. One is they’re starting to venture into a new business doing mobile phone tracking of drivers to improve real-time accident detection, things like that. Then the other line is core business analysis, like modeling what goes into their costs and how we can basically take advantage of their years and years of data to make better business decision. I’m kind of focused on that second one at the moment. It’s going to be a whole lot of regression analysis, machine learning and things of that nature that I’ve kind of always been on the outskirts of but never dove deep into. That’s what I’ve been doing for the last two months, just getting my feet wet so that I can contribute down that path. It’s been really interesting so far.
Matt: Okay, and so, in your day-to-day work, are you mostly doing programming? Is that what you spend most of your time on?
Scott: Yes, it’s programming but it’s more interactive than things I’ve done previously. I’m handed a big data set and I’m just kind of playing around and exploring with it in Python notebooks. It’s still programming, but it’s not like writing codes and then executing them, it’s more exploration at this point.
Matt: So, your product really is the results and the business intelligence you pull out of the data, not necessarily the code?
Matt: Okay. What motivated you to move to that from your former job at the defense contractor?
Scott: I did not like working for a defense contractor is the short answer.
Matt: What about it didn’t you like?
Scott: There’s a couple of things. One, it’s just so many more levels of bureaucracy than anything else I had ever been exposed to. I was working for a subcontractor, so we had to answer to this contractor, who had to answer to the government. There’s just so many levels of getting things approved and having to present, having to propose, to multiple levels to get work. We’re also a small company so we only had a couple of contracts. If one of them fell apart we were just kind of screwed as a division. Also, it sounds naïve, but just kind of the morality of it, working for a defense contractor. You’re working for military purposes and that’s just not something I was comfortable doing after a while.
Matt: Fair enough, I turned down a chance to work on nuclear weapons at one point for a fairly similar set of reasons.
Scott: I went into it knowing that that’s kind of how I felt, but what the company worked on is not the pointy end of the stick. We weren’t building weapons or anything, we were working on more of the surveillance type stuff. Going into it I was like, “maybe I can convince myself that that piece of it’s okay.” After about a year I had not succeeded at convincing myself.
Matt: I know for your dissertation you did a lot of stuff that you could broadly describe as data science or data analysis. When you were here with us at Tri Alpha Energy you had a very similar role. How would you describe transitioning from the sort of data analysis coding done for your dissertation versus out in the real world?
Scott: I had a couple of different experiences coding out in the real world. At Tri Alpha, as you know, it’s closer to an academic environment where we weren’t shipping code for product. If it got the job done and got you the right answer that’s really all anyone was concerned with for the most part. As a defense contractor, we were building software to be shipped and sold. There’s a huge difference when you need something to run to meet spec versus having code that just gets you the right results. It’s not even striking a balance so much as just your end goal has to change, and that was a significant transition for me. I wrote code when I was a grad student, I wrote codes that worked, but talk to people in my group, and they would probably say I was not a good coder.
I got the answers right but it was not the ideal way to get there. Working at Tri Alpha, I was kind of in that middle ground where it’s still academic but what I was doing there, we were writing code to run after every shot. It had to get done and it had to get done quickly. There’s a nice middle ground there where I was able to cut my teeth and actually become a much better coder. That prepared me for actually writing packageable code, which I was doing at Frontier Technology. It’s been a bit of a transition. Now I’m going back to more of the academic model where I just need to get the answers right. That’s been my experience. There’s those two elements throughout. Obviously, you need to get the answers right, but adding that extra layer of making everything efficient was the big difference between the academic and corporate coding.
Matt: Was it hard to pick up that new way of doing things or was it pretty easy and natural to make that transition?
Scott: There was a curve, obviously. The first couple of months of trying to basically change my mindset and making sure I was actually thinking about it where the hardest. Once I realized that that’s what I needed to do, I was able to make that motivation automatic, the mechanics of it was not very difficult. You know that that’s where you need to focus, you learn the tricks. You learn that string comparisons are slow, so you get rid of them as often as you can, things like that. You just kind of pick up the mechanics as you go and that piece of it’s pretty straight forward.
Matt: So let’s take it back a little bit. You did physics as an undergraduate, physics as a graduate student. Why did you decide to study physics in the first place?
Scott: I went into high school thinking I was going to be a math major when I got to college. I was always really good at math. I liked it. Basically, putting a math problem in front of me and having to get to the answer was what I enjoyed most about school. I was happy to go down that theoretical road until I took my first physics class and I was like wait, I can take math and apply it to real world, this is so much better. It was just really love at first sight in my 11th grade physics class, the first time I did a basic ballistics problem on where is the catapult projectile going to land. I was like all right, this is what I want to do and it stuck.
Matt: Cool. As you were going through your undergraduate and your graduate school did you have any sort of preconceived ideas about what your career after that would be like?
Scott: Absolutely zero. I’m still not sure.
Matt: You were in it for the physics and wherever that led you to a working job you were okay with I guess?
Scott: Exactly. Yeah, I went into undergrad not really wanting to enter a particular discipline. I worked in a lab for a couple of years and decided I didn’t want to do that. Then I was going to grad school and one of my requirements for grad school was a department that had a bunch of a different disciplines that I could go into because, even at that point, I didn’t know what I wanted to do. It’s always just been kind of let me get a wide range of options and I’ll kind of narrow it down from there.
Matt: Interesting and so after doing your dissertation, which was Search for a High Mass Electron-Muon Resonance I the ATLAS Detector at the LHC, you weren’t tempted to stay the course in high energy physics?
Scott: I was very tempted, I tried. The problem is, there’s kind of two problems that conflated themselves when I was finishing grad school. As I’m sure you know the LHC broke in 2008 and was down for about a year and half which caused a back log of grad students. Instead of the normal spread of grad students finishing and getting post docs, everyone working on the first year analysis finished at the same time. There was basically a little more than twice the normal volume. That made post docs that year incredibly competitive. That was one piece of it. The other piece of it was I didn’t actually want to continue working on the LHC. I wanted to switch to neutrino physics.
Switching disciplines between grad school and post doc is the hardest thing to convince people of, even within the focus of particle physics. There’s such a big difference between collider physics and neutrino physics that I found the going pretty tough. I spent like six months trying to get a post doc and failing. I had also been applying for corporate jobs and industry jobs at that same time. It just so happens that Tri Alpha came up on a random Monster search and it sounded very cool. It all worked out. I was being very picky with industry jobs. I was only searching in certain locations where I thought I wanted to live. I was only looking in certain industries. In the meantime, I was thinking, “Well, one of the neutrino labs will give me a job.” They didn’t. Tri Alpha did give me a job, so that’s where I went.
Matt: So far so good on that new course?
Scott: Yeah, I was very happy at Tri Alpha Energy. I was less happy at the defense contractor, but still happy with the overall decision to go industry and I’m happy now.
Matt: Excellent. I think, well we talked a little bit about what you see in your future. Do you have a particular career trajectory that you’re aiming for at this point?
Scott: Yeah, data science in general I think is where I like to be. It’s kind of goes back to what I was saying before, I like solving difficult math problems and data science is kind of the next phase of that. Getting into the machine learning pieces of it I think is something that I’m pretty interested in. I would not be surprised to see myself as a data scientist for the next long time.
Matt: Cool, and you think you’ll stick as a sort of an individual contributor or someone who’s directly involved in the data or do you have any thoughts of maybe trying to become management or leadership at some point?
Scott: Interestingly, at my interview one of the people I was going to end up working with asked me the standard question where do you see yourself in five years and my answer was a technical role, no interest in management, and that surprised him because people generally don’t say that apparently. That’s where I want to stay. Obviously, things change as circumstances change but as far as I’m concerned I want to be the boots on the ground for as long as I can.
Matt: That’s a good thing to know about yourself. As you know, I kind of took the opposite direction and it’s a different job. You definitely have to be sure that you’re interested in doing that different job of being a manager.
Matt: Okay, so last thing. Looking back on your time in graduate school, the transition we talked about, where you are now, what would be the one key piece of advice that sort of stands out to you that you would tell the folks ten years behind you who are maybe still in graduate school or thinking about going to graduate school?
Scott: That is a good question. I spent a little bit of time thinking about it earlier, but should’ve spent more, because I didn’t come up with a great answer. I think my grad school, they sent out a survey a couple of years ago asking grad student alumni what they, just about their experience, what they thought could’ve been added. Overwhelmingly, the response was career options other than academia need to be discussed. At least in my schools that never happened. Over the last couple years they’ve actually started having former grad students and other physicists in industry come in and give talks about that career, about that path. I think that’s positive. I think people get, especially in the early years of grad school, a lot of people end up kind of with a one track mind.
Okay, I’m going to go to grad school, I’m going to get my PhD to become a post doc to become a professor. For some people that’s the path. I think everyone goes into grad school thinking that that’s the path or at least a large percentage. It turns out that that’s not where the majority end up, obviously. I think the idea is that you have to start thinking about that early and start thinking about what it is that you actually want because there are ways to go through grad school where you love what you do, but you know after the five, six, seven years of your PhD you’re done with it and you want to move on. If you can start thinking about those alternate paths earlier, then I think that you’ll end up making a better decision. You’re not going to be stressed two months after your defense, like “I can’t figure out what I want to do” and so “I’m just going to go teach.”
That’s where a lot of my friends ended up and they’re in teaching-only positions because that was the only thing they could do with their PhD at the end. It’s the only thing they could do quickly at the end of the PhD. I think just kind of thinking about alternatives is something people need to do earlier unless they’re 100% “I’m going to get a post doc and I don’t care if it takes me a year to do it.” If you’re that committed, good luck to you.
Matt: Usually, you also have to be willing to go anywhere on the planet as well.
Scott: Exactly. Yeah, I mean I don’t know because I didn’t apply for them. I think I could have gotten a post doc if I was willing to live in France again. I just wasn’t.
Matt: Fair enough.
Scott: I think that’s where I’d go with the advice question.
Matt: Okay. Well, Scott, thanks again for your time. I really appreciate you sharing some of your experiences with us. I wish you all the best in your future as a data scientist.
Scott: Thank you, Matt