Alumni Q&A: Sean Wade on Solving Healthcare Problems with Data Science
Based on July 2020 interview
Sean Wade
ACME Class of 2017
Apple, Data Scientist
Tech, Healthcare, Wearables
Highlights
- Data science—especially healthcare data science—requires an every-day commitment to ethics.
- Take time to get to know people—an offhanded introduction ultimately to an interview and job.
- Stay above water with ACME, focus on finding what about it you enjoy, and pursue that.
Tim Riser: Can you share a little about yourself your background and where you’re at right now?
Sean Wade: I graduated ACME in 2017 and after that got my masters at BYU in Computer Science. This past December, six or seven months ago, I graduated and got a job here at Apple. I’m doing data science for the Health Strategic Initiatives. We do lots of really cool work with the Apple Watch and medical data to help people to live healthy, happier lives. It’s a cool space and Apple has some of the best data you could hope to work with.
Tim Riser: When did you decide to do ACME, and what were you considering when you made that choice?
Sean Wade: It was fortunately pretty early on. My first semester I made a list of majors I thought were interesting. I was into physics—that was probably my first one. I also looked at electrical engineering, but I honestly never considered math, until I saw one of the little signs that ACME had. I thought, “That’s interesting.” I was taking my first programming class and it seemed related, so I decided to see what ACME was. I talked to Jeff and he was very convincing. Seemed like the future and I was in.
Tim Riser: You’re at Apple after finishing your master’s degree. What drew you to Apple and this opportunity?
Sean Wade: I knew I wanted to do a data science-type role. I had done software engineering before and even some machine learning engineering work. However, I really enjoyed the higher level perspective of data science, where you ask questions like, “What are our limitations? What partners do we work with? How do we make meaningful change?” The data scientist role had those traits, and that’s what drew me to it, particularly this one at Apple. I’ve done some work in the past in wearables and health, and Apple is the leader in terms of that. I spoke to my future manager, and he said they have programs with hundreds of thousands of people and there’s no other health data set with that kind of fidelity, or with that kind of impact where you push changes and it would impact their lives immediately. Apple drew me in with both the fun problems and the real-world impact of our choices.
Tim Riser: You talked about a couple different roles. I think it’s helpful to articulate the distinctions that you see very clearly that you might not have in ACME. Talk to me about ML engineer, data analyst, software engineer, data scientist. Can you describe a little bit about what you’ve done and how those worlds are distinct?
Sean Wade: Yeah, I think they’re very fuzzy terms—especially when you get to mid-range and smaller companies. They’re very interchangeable. I worked at a start-up where I was a ML engineer, but that involved pure software engineering, data pipelines, building the model—it was really everything. But once you get to Google, Apple, and the bigger companies, it’s more defined. Machine learning’s primary focus is building models that can be deployed into production. That could be translation or image recognition or things like that. A software engineer could kind of do some of that, but they’ll do more of the pipeline. A research scientist focuses more on the idea of developing propositions, seeing if it works on a small scale and as a minimum viable product. A data scientist works with products and data to improve it and drive change.
Tim Riser: You think there’s anything ACME students might commonly misunderstand about the data scientist role?
Sean Wade: One of my big misunderstandings from ACME is the belief that it’s all about hard math, optimization, and state-of-the-art models. While that’s sometimes true for data science, the most important thing is to have a strong foundation on the fundamentals and be able to simply frame a question. Know the basic methods like regression and then understand the assumptions and circumstances where you know you need to go deeper and employ more advanced models. Everything is a trade off with accuracy, explainability, your time and privacy constraints. Meaningful impact is not driven from the cool state-of-the-art, but rather from getting things done, aligning with management, and showing actual results.
Tim Riser: Did you do the master’s degree with an idea of going into industry?
Sean Wade: I knew for sure I wanted to do industry. I wasn’t sure if I wanted to do a PhD. I was close with a professor I had some classes with, and they were already doing some research. I decided to try it out with him, work on a thesis, and see if it’s for me. I did it and liked it, but I knew I didn’t want to be a pure researcher. I was at the point in my development where I could work on and lead interesting projects so more school wasn’t necessary for that goal.
Tim Riser: Once you had the skill set where you felt you could make it, how did you go about getting a full-time role in industry?
Sean Wade: Internships were probably the biggest thing and the second was making connections and talking to people. My first internship was at a local eye tracking company. It was nothing super fancy, but it was math and programming experience. From there each internship and job built of the last and got better. The next summer I did MIT Lincoln Lab and then AI at Microsoft. They all built on each other, to the point where my resume started having a diverse set of experiences. And so when I went into my current role, one of the hiring decisions that I found out after was that they were impressed with my breadth out of college. That was something that helped them make the hiring decision, even though I lacked a stats PhD or years of job experience. They saw that I filled a niche where I could learn the other skills necessary, but I had a wide and solid base. The second thing was meeting great people. I sent out a bunch of cold emails and things like that. I think everyone knows thats very hard, and there’s lots of rejection there. I had a good friend from ACME, Rex, who worked at Apple and he had met this guy and offhandedly introduced us. That’s how I ended up meeting my manager and how I got the interview that ultimately got me the job. Getting a job is a lot of luck but internships and developing great relationships made it possible for me.
Tim Riser: By the way, what was the program you did?
Sean Wade: I did CS with David Wingate.
Tim Riser: If someone’s trying to make the decision whether to go straight into industry out of undergrad and get an associate data science or data analyst role and work their way up, or first get a master’s and the try and jump into industry, what would your contribution to that decision be?
Sean Wade: It depends I think on the role, but I’ll take it from a data scientist perspective. At the bigger tech companies most positions prefer PhDs, then some master’s students and since they have tons of applicants. The less education you have the more you have to really stand out in other areas, like research papers or work experience. Since data scientist is such an overused term it really helps to signal you have a strong academic background as well and not just a 9 week bootcamp. After I got my masters I noticed I received much more callbacks on my data scientist applications than right out of the bachelors. That route worked out well for me and I would probably recommend at least a masters, but there are tons of people smarter than I am who have awesome open source projects, work experience, or research that can get them the interview with just a BS.
Tim Riser: What was the skills step up in your master’s degree, beyond the degree itself making you more employable? What did it add to ACME?
Sean Wade: That’s important. It wasn’t an academic course load where I really felt like I was learning a ton of new material. That was more during ACME than in my master’s program. My master’s program was important for two reasons. The first one is the attitude of “This is all on you. This is your research. You really direct everything.” The classes are thinking of a problem, narrowing the scope, and solving it. From that aspect, the method of problem solving is much more applicable to business, because the main lesson is how do you deal with ambiguity and how do you use what you learned in ACME to actually solve a problem. It’s very different when a problem is given to you versus created by you. The second one was that my master’s program gave me a broader perspective. As I was trying to solve these problems I realized we went really deep in some things in ACME, but other things were more surface level. Grad school helped me solidify those foundations by putting them into practice.
Tim Riser: You’ve been in the “real world”, interacted with colleagues, seen a lot of big deal products. How would you say ACME prepared you?
Sean Wade: Across some dimensions, I would say extremely well. On the dimensions of teamwork and collaboration, ACME does a very good job by forcing you to do that. I’d rank it very high in terms of basic stats and like these other things that act me that I didn’t really get too deep in Acme that I had to learn outside of it. It made me aware. Once I got to the job, I realized that just because I’m aware of these things doesn’t mean I know them in practice. Doing these things like the back of your hand like you will need to won’t come automatically with your ACME degree.
Tim Riser: Are there any topics ACME didn’t cover that would have helped you with your job?
Sean Wade: I really appreciate ACME because it was the only place that I knew at BYU that offers such a broad sweep of many fields and lots of different mathematical tools. To what I said before, my one weakness coming out of ACME was that I knew some basic stats and some experimental design but I was weaker than other stats candidates because I had never gone through the “times tables equivalent” of stats, where you have the knowledge cached. It was one of the things I had to work on after.
Tim Riser: If you were to do it again, would you do ACME?
Sean Wade: Absolutely. If you were to ask me at the beginning of year four maybe I would have been more shaky, but after coming out, yes.
Tim Riser: That’s good benchmarking. Great to get time series data! One of the things I got out of doing strategy consulting was an appreciation for the snowflake-ness of every company. Each is built differently, has its own unique organization, characteristics, and so on. I’d love to hear your thoughts about your current data science team.
Sean Wade: Apple is very interesting. Like you said about the snowflake analogy, it’s a privacy-first company and it is also a product-first company. So unlike Netflix, Facebook, and Google that really evolved with data in mind. Apple evolved out of a product and data was tacked on second. Data wasn’t built into the DNA from the start. Because of that, you have a different experience. I have colleagues at Facebook or Google who have all this infrastructure built out where they can run all these tests at huge scale. At Apple, we actively turn away data because we value privacy. That’s our selling point. Since we are limited on the data, we have to be creative, and do federated learning, on device modeling and things like that. Since my health team is relatively young there is lots of change and growth rather than existing pipelines. Because of that we work closely with data engineers and many other software engineers for our infrastructure. At the same time, we work very closely with everything from UX designers to doctors. I work with a lot of different people, and that level of an interdisciplinary experience is, I think, unique to Apple.
Tim Riser: Data science is a really interesting toolkit for viewing and interacting with the world, but as often happens when we’re developing, teaching, and learning a toolkit, discussions around appropriate uses of the toolkit, or toolkit ethics, often come second. How did that focus on privacy factor into your decision to join Apple? And what have you learned about ethics through your work?
Sean Wade: I was always privacy conscious, but Apple takes privacy much more seriously than everywhere else I have worked. It has been a learning experience for me. Every project and piece of data we touch has to be validated, and we have to say exactly what the business cases are and justify how we are using it. Whereas many data scientists and even me coming out of school would say, “Give me as much data as possible, and I’ll sift through it, get the model, and then check for biases,” Apple says: “Tell us what you’re going to do before, and then we’re going to check for biases before you build models, and we’re going to put privacy constraints on the data before you see it. We’re going to have data you can’t even touch, but you can interact with in some way.” We’re privacy focused. Especially working in health data, there are additional regulations and you really have to be cautious about everything you do. To your point about ethics, I have learned that, even with the best intentions, it is very easy to have unintended consequences with algorithms. Models can disproportionately effect populations by age, weight, sex, race and just about anything else. As data scientists it is our job to actively test and avoid this.
Tim Riser: We cover a lot of fundamentals in ACME. What are the fundamentals of ethics when it comes to data science?
Sean Wade: Yeah, I think it all boils down to asking what does your model or your algorithm touch? And if it’s touching something as important as health or medical records, then you really have to dedicate conscious time to it. If it’s touching who gets a loan, if it’s touching who gets hired for jobs, and these sorts of things, before you train, during, and after, you need to have fairness and ethics in mind. Ethics is thinking about it through the whole process, not just at the end.
Tim Riser: So we started this conversational thread off talking about the structure of your team. What are the other teams that you interact with?
Sean Wade: Yeah, that is one thing I really do like about being a data scientist. As a software engineer in my experience you kind of get siloed and you only work with other engineers and sometimes management. As a data scientist we get to shape how data and the products we work on are thought of, built, shipped, and used. Every day I get to work closely with designers, medical doctors, researchers, data engineers, and management. A huge part of data science is working with other people and communicating mathy things to non-mathy people.
Tim Riser: What final thoughts would you have around choosing where to spend your time and energy during ACME to prepare for a successful career and successful entry into industry.
Sean Wade: For me, the biggest thing that I did to impact my trajectory was to stay above water in ACME, really do the basics, and then not worry so much about ACME that I couldn’t do other things. Reaching out for internships, that was a huge thing I did. Trying to get a broad range of experience and background was another thing that really helped me. ACME was really helpful in that regard because it set up meetings where people came to talk to us. I was super-bonded with everyone in my class because we spent every waking hour together. That was very helpful in life—talking to them about what they’re interested in and getting referrals to other jobs from them. If I had to go back and tell myself anything, I’d say, “Stay above water. Do the material as much as you can, but really focus on finding what about math you enjoy and pursuing that.”
Tim Riser: That’s my favorite analogy I’ve heard: “Just stay above water.” It’s a hard balance.
Sean Wade: On all my assignments I could have worked longer, but there’s a point of diminishing returns. You have to know when you’ve learned the material and how to cut your losses on some of the things to focus on others, right?
Tim Riser: My final question is, what kind of life advice would you have for today’s ACME students?
Sean Wade: I don’t know if I’m qualified for any of that. In ACME, you’re surrounded by overachievers from the start. Your sample is biased, so it’s easy to look around and see people who are smarter than you, who have better opportunities than you, who have better jobs than you. As I’ve gone along, I’ve realized that as long as you keep progressing, you are good to the people around you, and you try to do things that you’re interested in, it kind of all works out in the end. I never had this end in mind, but I did smaller, less intimidating steps. Meet cool people and do interesting things. That will put you in a great direction. So it’s important to just keep going, no matter where you’re at.
Tim Riser: Thank you Sean. Thanks for taking the time to do this.