CDS Incredible Alumni: Seda Bilaloglu

At CDS we aim to give our students the skills and knowledge to go out into the world and innovate, create change, and make new discoveries. CDS students go on to do just that in industries and locations all over the world. To celebrate these achievements we like to introduce these students to you here! So without further ado, we are proud to introduce one of CDS’s Incredible Alumni, Seda Bilaloglu!

Seda graduated from CDS this Spring and is currently a Data Scientist/Engineer at NYU’s Department of Population Health. Recently she published a paper on thrombosis in hospitalized COVID-19 patients. We asked Seda about how she came to data science and how data science can help combat the COVID-19 pandemic.

Could you tell us what drew you to studying data science and eventually, CDS?

Since I was young, I knew I wanted to pursue a career in technology, as I come from a family of engineers. I studied Electrical Engineering for my undergrad and afterwards earned a master’s degree in Biomedical Engineering at NYU Tandon. Later, I started to do research in Motor Recovery within the Rehabilitation Medicine Department (at Rusk Rehabilitation). I was mainly specialized in digital signal processing and statistical analysis and was exposed to developing technologies such as wearables and telemedicine. One of my first data science projects was building clustering algorithms to find patient subgroups that responded to a given therapy. During this time in 2015, machine learning and big data became extremely hot topics. I then came across CDS and got very excited to one day be part of all of the interesting work done by the faculty and alumni. NYU Langone was a big help with employee tuition benefits, which made my decision much easier as I eventually applied and got accepted, and very glad I did.

What was your experience at CDS like?

It was really great; I highly recommend it. I did the program part-time for four years. It was challenging at times to keep up with a full-time job and all the coursework but it was quite manageable after making use of the office hours and working with my classmates. The courses were well-designed and focused on our understanding of the concepts. The faculty and staff were very supportive and open.

Aside from the classes, there was a good amount of fun, engaging social activities, and happy hours where we got to spend time with others in the program. I met very good friends that I am still in touch with.

What have you been working on since graduating?

Prior to COVID-19, I was working on predicting mortality from clinical text and appointment no-shows from Electronic Health Records. However, since early March, I have been mainly focused on predictive modeling, reporting and analysis for COVID patients.

I mainly employed classical Machine Learning techniques or deep learning models within these works and covered all assets including engineering tasks like data querying, cleaning, modeling, experimentation and deployment.

You recently published a paper on the effects of COVID-19 on Hospitalized patients. How did you get involved in COVID-19 research?

In early March, my team, the Predictive Analytic Unit, directed all its focus to COVID related operational tasks and research. I was assigned to my specific project– extracting thrombosis events from radiology reports–because I had prior NLP experience. I then started to work with a group of medical students to identify the clinical definitions and tuned the medical concept extraction tool for the best performance. We performed many chart reviews and constantly received feedback from the clinical experts. After extracting the thrombosis events, we employed competing risk analysis methods to understand the relationship between various covariates and thrombosis. This work was then published in JAMA. It was a very intense and fast paced couple of months, but was very rewarding in the end.

How do you feel that data science can help fight the pandemic?

There are multiple ways in which data science can help fight the pandemic. One way is to provide accurate estimates of health care demands so that they can be kept within manageable levels. Data can also help pinpoint outbreaks and predict spread, contact tracing, and transmission. Another way is to build clinical support tools that help clinical staff carry out diagnoses and get early warnings for admitted patients or those visiting the emergency department. Various data and skill sets can be used for these kinds of applications: medical texts, images, and electronic health records etc.