Meet the Team Series — Michelle Ye edition

When considering a new job, people wonder, “Who will I work with? How will I be able to grow with them? Do I want to work with them?” We have designed our interview process to help answer these questions among others. But before you join us at Arena, we’d like to provide you with another way to meet us: the “Meet the Team” blog series, where we have a conversation with various current employees. Below is our first with the Vice President of Data Science, Michelle Ye.

What do you do at Arena?

Michelle Ye: At Arena I lead the Data Science team, which is responsible for our core model development and refinement, in addition to some analytics tasks.

How did you get into the field? (I’d like you to talk about your technical experience and how your past path led you to where you are today)

MY: I came from an economic theory background where, as a math-based subfield, insight and parsimony are highly valued, but not real-life applications. At some point I realized that although I prefer the elegance of theories, empirical work leads to real life impact, which I value more.

Motivated by the prospect of improving the quality of real lives, I left academia — where I did some teaching and academic mentoring in addition to research — and joined Bermuda Health Council — a specialized quango spearheading the national healthcare reform in Bermuda — to lead their analytics arm. The national healthcare reform was an overhaul of the entire healthcare system, in scope and depth more ambitious than any healthcare reform in the past decades in the US, and much analysis was needed to understand potential legislative or policy changes.

After Bermuda I decided to move to New York, and that’s when I joined Zocdoc, an online platform for booking doctors’ appointments. There, based on my observation that patients usually are not trained in medical jargon and taxonomy and can’t be expected to know which medical specialist they need to see before they can book an appointment, as was required by the UI at the time, I wrote a smart-search algorithm, which takes any user query and returns the relevant medical specialties. This makes it easier for prospective patients to find suitable doctors quickly and is currently live on Zocdoc’s website.

I learned a lot on these jobs. In Bermuda, it was particularly inspiring to meet many highly competent people based in the public sector who have devoted their careers to the betterment of their community, instead of focusing on other career choices that may, for example, provide more financial reward. Interestingly enough, while on that job I also started to notice the benefits of theory training, because we were routinely required to solve ambiguous, sometimes messy analytics problems. Insight and parsimony are useful because they make an ambiguous problem solvable and lead to approaches that are more likely to be convincing to both decision-makers and those affected by these decisions. Data science is not a passive collection of tools and mindless applications of known procedures; rather, the practice of data science needs to be guided by theory and intuition. This is also why good data science teams often are composed of data scientists from varied backgrounds. Those differences in background training frequently result in different preferences for techniques and hone different intuitions, and the team can complement each other in solving the jigsaw puzzle of data science.

What problem at Arena are you most excited about solving?

MY: The most fundamental challenge for me and my team is to effectively use data science to improve workforce outcomes, in concert with the Product team. Considering we are a startup, I generally take solving all types of business problems as my mandate. Sometimes I take ownership directly, but I always encourage others to share their take on solving problems (regardless of whether these problems are from their own functional areas) and at times limit myself to only contributing suggestions. As for specifics, I trust if you ask me every month my answer would be different, but I’m most excited when I can make a significant improvement in any area of the business. At any startup, there are always technical, organizational and cultural challenges as the startup finds its way and its voice in the world.

Why did you join Arena?

MY: Mainly for the challenge. At Arena we solve very challenging data science problems with no known blueprint. Solving these problems could lead to significantly positive social impact if we as a company execute on our vision well. Indeed I feel fortunate to be at this company and have the opportunity to bring in such impact.

What topics/ideas/concepts are you concentrating on learning right now?

MY: I generally follow new research and industry trends. One cannot stop learning; as a data scientist particularly so. And this is a golden age for learning data science, since it is an active field generating exciting new discoveries daily. I recall that in the year 1997, IBM’s Deep Blue beat the world chess champion Gary Kasparov. A machine beating a human at an intellectual task, it was an event that shocked the world and made an indelible impression on me, and I recall being so curious to know how Deep Blue did that. But I had no way of finding out at that time; in fact, I couldn’t even find a popular explanation that provided any clarity. Fast forward to recent years, when DeepMind’s AlphaGo beat the world’s best Go players, I can perform a simple search and get and read the paper that explains just how AlphaGo does that. On top of that, a whole community of data scientists were able to exchange their interpretations of AlphaGo’s approach,relate them to past experimentation and then make suggestions. Soon after that, Deep Mind was able to publish another paper using a much more elegant approach that reliably defeats its previous algorithm. Contrast this with what happened in 1997, our current level of knowledge spillover in the data science community and the speed of progress is simply incredible. It’s impossible for me not to be drawn to the advances in data science.