Edmund Sumpena is a junior completing a Bachelors of Science in Computer Science and Neuroscience. He is a recipient of the 2025 Barry M. Goldwater Scholarship. In an interview with The News-Letter, Sumpena described his research experience and future plans.
The News-Letter: Could you give us a brief description of your research projects?
Edmund Sumpena: I joined Craig Jones and Amir Kashani’s lab [in] freshman year and started leading a large-scale project with multiple parts that has spanned until now. I'm developing AI technology that has translational and clinical implications, which includes developing generative models, developing convolutional neural networks and making sure that it outputs images that are clinically interpretable and reliable for diagnostic purposes and clinical studies. Specifically, I’m working on improving clinical workflows involving quality assessment and screening in optical coherence tomography angiography (OCTA). These are 3D, high-resolution images of the retina. The first part was essentially doing artifact detection: recognizing quality defects that are part of imaging and ensuring good quality. And the second part, which is what I'm doing now, is to improve the quality of images without adding in fake vessel structures as part of the images.
N-L: What tools are you using in your research?
ES: Our project is really unique in the sense that it is not really a topic that's been explored very extensively in literature. We’re working on high-resolution OCTA imaging data. And in a lot of literature, they focus on 2D images, but our lab has been really focused on doing a more thorough and comprehensive analysis on 3D volumes. So that means I'm creating AI models from scratch on PyTorch — an AI development platform — designing and implementing ideas from scratch, and I've come to realize that that just takes way longer [than repurposing previously existing models]. But I think it's just been a really rewarding process of kind of learning through application and figuring out exactly how to develop models that are clinically relevant and applicable.
N-L: What drew you to your particular research field?
ES: One of the mobile apps I developed in high school was geofencing technology to detect when people exit or enter a particular region. I applied this technology to help dementia patients who are wandering, where a patient leaves a facility and then could go outside and get injured. I ended up taking my app and developing it into an anti-wandering technology app called WanderGuardian and tested it on real people with dementia. The process of going to them, figuring out what I need to improve and continuing to refine the app, taught me that the value in developing technology is in its application. And that drew me into medical AI research in general, which brought me to the Wilmer Eye Institute. We're working with direct clinical data, not for the sole purpose of publishing a paper or creating a shiny new model, but to work with different centers to apply our technology.
N-L: How has Hopkins and its programs supported your independence and ability to explore your research topics?
ES: One of the main reasons why I came to Hopkins was because of the interdisciplinary emphasis between engineering and medicine. One thing that I really appreciate about Kashani and Jones is that they give me a lot of independence in the way I'm developing my technology. I've learned the ins and outs of the research process, starting from developing a new idea to project proposals and grants to developing the technology, and then finally writing out the manuscript. I think that it's been a valuable experience.
The professors are really supportive. I also think that with Hopkins, there's a lot of flexibility in what path you want to take in terms of your course progression. I took some in-depth upper-level [computer science] classes in my early stages, like Natural Language Processing, and I also took Cells and Systems, which gives me a broad, diverse set of classes. And I think that as that happens, it allows me to build upon all that foundational knowledge. It also allowed me to build the toolkit in order to continue my research. Jones and Kashani allow me to break things, fail and then figure out what failed. They not only give me lots of support and guidance but also an experience on how to fail firsthand so that I can learn to tackle the obstacles from that. What I enjoy about research is to figure out how to innovate.
And for extracurricular support: one club I'm involved in is Advocates for Baltimore Community Health, a community service club where you go outside, volunteer and interact with the homeless. We do clinical volunteering, we help the disabled, and being able to interact with those people and understanding who we're helping and visualizing who we're helping. Seeing that so many people have ongoing, unmet needs has grounded me in the bigger picture. It’s reminded me that my work isn't just about the technical details of a specific project: It's about serving society and improving the well-being of the community. That broader sense of purpose has been a major motivating force for me here at Hopkins.
N-L: What’s your ideation process like?
ES: It's certainly a lot of reading of existing literature, but also relying a lot on the experts that we have, like the ophthalmologists. And one thing that I've found to be the bulk of research is that I start off with a baseline idea, a baseline concept, and it's amazing how one problem branches off into a ton of separate problems. If I believed in my idea, I would continue to develop it, and then spend a couple months working on it. One of my ideas was adopted and became one of the main parts of our project. Ideation of new research projects requires curiosity and the willingness to innovate and explore.
N-L: Were there any challenges you faced during your research, and how did you overcome them?
ES: I think the hardest part about research is I’m tackling a problem that no one else has tackled. There are so many times where I encounter a problem, which I don't know how to solve, I spend a lot of time just to sit and think about the problem. I think the value of simply thinking is often underrated, especially as a researcher or student, where there’s constant pressure to always be doing or trying something new. But taking a moment to pause and reflect can be incredibly powerful. Sometimes the best ideas come from exploring connections to fields that aren't directly related to medicine, like physics or neuroscience. Drawing inspiration from these areas can offer fresh perspectives and help move a project or even the entire field forward.
For example, in my first project, we were having a lot of issues because optical scans contain a lot of fine vasculature. These are high-frequency signals. It's very easy for convolutional neural networks, which is a type of AI model, to treat them as noise. Essentially, they're designed to filter out noise as part of its learning process. So I thought long and hard about this. The model had to attend to find details, but also “understand” the bigger picture. Neural networks can't do this inherently. So I ended up developing a model that allows each layer of the neural network to essentially integrate information from the fine details but also the global structures. And that was a game changer in that project.
N-L: What advice do you have for students looking to apply for the scholarship?
I'd say the biggest advice I have for students is to be genuine about your motivation. Don't try to craft your application based on what you think they like. Craft it based on your genuine motivation and drive behind your research and talk about all the experiences where you took initiative in your research. Talk about the experiences where you get hands-on experience with grant writing and project goals; different dimensions of research.
