Andrea Clark-Sevilla reveals how taking a master’s allowed her time to figure out if she wanted to pursue a PhD and research career.
It’s a Friday night at International House and the graduate student residents of the dormitory are gathered to watch Central do Brasil. Andrea Clark-Sevilla is among them and looks forward to immersing herself in the touching story about friendship and finding one’s greater purpose. It is a much-needed break from the busy last semester of her master’s degree.
The past two years have been non-stop for Clark even though she started her graduate career in 2020 at the height of the pandemic. Although she was fully remote, living in Querétaro, Mexico during her first year, she managed to pack in a research project with Senior Lecturer Ansaf Salleb-Aouissi and win a National Institutes of Health contest for it.
The Decoding Maternal Morbidity Data Challenge aims to promote and advance research on pregnancy and maternal health. For the challenge, the team looked at data from the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-be (nuMoM2b) and decided to focus on preeclampsia, a pregnancy complication tied to high blood pressure that could lead to maternal and infant death if left untreated. Together with colleagues from Hunter College, their research, On Predicting and Understanding Preeclampsia: A Machine Learning Approach, developed a machine learning model that can predict women at risk of developing preeclampsia.
This was Clark’s first research project and being part of it helped her decide to pursue a PhD. Clark shared that she looked at the opportunity to use the two-year master’s program to explore and see if a research career is for her. “There are so many different ways to do research and chances to do other interesting things at Columbia, that I was hooked!” said Clark, she will begin her PhD in the fall and continue working with Salleb-Aouissi. “I now know for sure that I want to become a researcher and I am looking forward to starting my PhD.”
While many think that having to do meetings over Zoom and not being able to work and collaborate in person is a hindrance to good work, the opposite is true for Clark. Over the summer, she was able to do an internship with the Johns Hopkins University Applied Physics Laboratory (APL) and volunteered as a student instructor at Columbia’s AI4All summer program for high school students, where she was invited by Professor Augustin Chaintreau to lead a class on machine learning. So while doing research remotely from El Paso, Texas she would jump onto Zoom sessions with the AI4All students who were in New York City. Clark admitted she would not have been able to do both if things were in person.
When she isn’t brushing up on her French and German skills or watching a foreign film or two, Clark is working on her final projects and schoolwork and attending meetings to write the research paper for the preeclampsia project. We sat down with Clark to find out more about how she decided to pursue a PhD and her new love of research.
Q: Did you always want to do research and how did you start working with Professor Ansaf Salleb-Aouissi?
I studied math as an undergrad at Cornell, and it was not until rather late in my program that I found out what application field really interested me research-wise. I took a course in dynamical systems and biology, and it was after this that I found that I was passionate about combining biology and computing.
I actually found Professor Salleb-Aouissi through her wonderful and engaging edX course on Artificial Intelligence. After looking into her research areas, I was absolutely convinced that I wanted to work with her and applied to Columbia hoping to get a chance to collaborate with her in some capacity. It was so incredibly fortunate that it happened to work out that she was looking for students to work on a project during my second semester in the program!
Q: What did you work on and what did you like about the research?
I have mostly worked with evaluating different existing methods for interpreting traditional black-box machine learning models. For the NIH challenge, I leveraged a method called Partial Dependence Plots (PDPs) to determine which feature(s) had the greatest marginal contribution to a model we trained for predicting the incidence of preeclampsia in pregnant women. Using this method, we were able to narrow the cut-off points for high-risk factors, such as body mass index (BMI), blood pressure, and some notable placental analytes (proteins and/or hormones generated by the placenta) and show their influence on the model’s ability to predict the incidence of preeclampsia.
This can be useful information to clinicians who wish to monitor their patients based on a more curated set of risk factors and critical ranges for these, as well as organizations such as the American College of Obstetricians and Gynecologists (ACOG) who largely set the guidelines for this medical evaluation. Drafting appropriate guiding criteria for such a potentially dangerous condition has the potential to save many women.
Q: What are you working on now?
We are currently preparing the paper related to the NIH challenge for publication. Our team that worked on this challenge had so many great ideas, and the paper is slowly but surely evolving to its final form. The challenge itself felt rather short-lived, given how rich the data is and the different angles to approach the problem, depending on how one defines preeclampsia, for instance. All these details need to be properly addressed and defended, which takes much time. I am also finishing up my course requirements to graduate.
Q: Why did you decide to get a master’s degree instead of applying for a PhD?
For me, it was very important to figure out if I was suited to doing research before committing to a five-year program in which I would be doing this exclusively. I have heard stories about students dropping out of their PhD programs because it was not what they were thinking they signed up for. I don’t think the traditional undergraduate curriculum adequately prepares one for research, or at least it was not the case for me.
Q: What do you think people should know about doing a master’s degree? If you didn’t go through the program would you have applied for a PhD program?
A master’s degree is a great option if your undergraduate degree is not well-aligned with your career objectives, as it can give you the opportunity to pivot your skillset accordingly. I would say my experience is not the normal use-case for it, as I purely pursued the master’s degree to decide if I enjoyed doing research and could see myself continuing in a PhD program. If I did not enjoy it at all, two years is not a lot of sacrifice career-wise, and it is certainly a good learning experience.
The same cannot be said for a PhD program. I personally would not have had the confidence to commit to a PhD program not having the research experience I had with Professor Salleb-Aouissi. It is a bit of a double risk with a PhD program. First, you need to be reasonably committed to your research topic, and second, and I think most importantly, you need to be confident that you can work well with your advisor.
I feel that many students go into a PhD blindly, straight after undergrad. I am extremely fortunate to be able to say that I am very confident about both thanks to my experience in the master’s program.
I was also lucky enough to win a National GEM Consortium Fellowship for the master’s program. The fellowship allowed me to focus primarily on my research and not have to worry about the financial aspects of being in the program. I was also awarded this fellowship to continue funding my PhD studies.
Q: Why did you decide to apply for a PhD?
I feel that I am very driven when I have control over the questions that I want to answer and I have the freedom to explore them in the ways I see fit. I think that the most suited profession for someone with these characteristics is research.
Doing a PhD gives you the freedom to live in your own little world for five years and come out an expert in what you are passionate about. It’s really a dream situation!
Q: What will be your research focus?
I will continue my research on explainable artificial intelligence, likely in the precision medicine field. I also hope to be able to dig more into the theoretical underpinnings of more statistics-driven approaches and develop my own approaches for interpreting machine learning models.
Q: What sort of research questions or issues do you hope to answer?
I would like to bring the issue of creating explainable AI more to the forefront in the machine learning community. I feel that there is a lot more focus on developing the most state-of-the-art models in terms of predictive performance, but there is not enough research being done to make the results of such models understandable to the end-user, which might very well have serious social impacts.
Q: What is your advice to students on how to navigate their time at Columbia? If they want to do research what should they know or do to prepare?
I think students should take classes that truly interest them, and if possible, also explore courses in other related departments. I have a friend who is taking a project-based course combining data science and climate change, and he is learning so much from it and enjoying it greatly!
I personally think that the best incarnation of learning comes from working on a tangible project like that and having the space to try ideas and explore. That is how it started with me and Professor Salleb-Aouissi.
Q: Is there anything else that you think people should know?
Don’t be afraid to fail at something at first! I always felt pressured to get something right on the first attempt, and I quickly realized that this mentality is not sustainable in the long run if you do research.
You learn so much more from your mistakes than from your successes. Your critical thinking skills are actively engaged when you have to analyze why something failed as compared to when it happily worked on the first try. I would hazard to say that researchers are skilled puzzlers because they always manage to pick up the pieces when something breaks.