I am a PhD student in Medical Engineering and Medical Physics at the Harvard-MIT Program, concentrating in Computer Science. I aim to create machine learning technologies that help medical professionals provide the best possible care to patients, especially in low-resource settings.

Prior to starting my Ph.D., I obtained my B.S. in Biomedical Engineering (minor in Computer Science) from Bucknell University. I was fortunate to work with Ben Wheatley on computational biomechanics, with Joshua Stough on machine learning-based healthcare, and with Aalpen Patel on liver tumor ablation probes. Outside of research, I also worked at Livongo Health (now Teladoc Health) as a backend software developer and helped create micro-services monitoring API. At Livongo, I had the chance to work Norman Chen, Kunal Sachdeva, and Suresh Margabandu.

Research Highlights

During my PhD, I am primarily interested in exploring questions such as:

• Multi-modal data fusion: How can we inelligently combine data from various clinical modalities (ex: histology images, spatial transcriptomics, and EHR) to make clinical workflows more efficient and healthcare more personalized?
• Machine learning for novel imaging technologies: How can we solve the theoretical challenges associated with using machine learning for novel imaging technologies, like 3D pathology?

Prior to my doctoral studies, my research experiences included:

• Clinical Decision Support: Many clinical decision support systems suffer from intra-observer variability, workflow bottlenecks, and high costs, limiting the reach of healthcare systems. Automated diagnosis systems leveraging large amounts of data are a key solution. Since starting out in research, I have worked on various projects, including 1) Severity classification of diabetic retinopathy in fundus photographs and 2) Optimizing transfer learning for COVID-19 diagnosis in small CT-scans datasets.
• Domain Transfer and Generative Modeling: Many African countries have a single MRI machine for a million patients, thus underscoring the need for technologis that will reduce MRI acquisition times and costs. Since an MRI sequence often consists of T1-weighted and T2-weighted scans, can we use machine learning to translate a patient’s T1 scan into their T2 scan, thus cutting down on MRI times? In my previous work, I showed how incorporating stylistic losses in generative adversarial networks can help translate both healthy and unhealthy T1 brain MRI to T2 scans.
• Tumor Ablation Technologies: Liver tumor ablation is a minimally-invasive procedure to kill tumors in situ by heating them. However, current technologies only allow ablation of spherical tumors. Along with my senior capstone team, I designed an ablation probe that provides greater spatial control to interventional radiologists during tumor ablation procedures.
• Computational Biomechanics: Numerous crash testing and surgical planning softwares rely on accuracte computational models of skeletal muscles. However, the role of fluid in compressive skeletal muscle biomechanics as well as other surrounding tissues like the aponeurosis is not fully understood. In my undergraduate research, I explored viscoelastic and finite element models of skeletal muscle and aponeurosis to understand the role of fluid in their mechanical properties.

Aside from these projects, I have also ventured into:

• Predicting Cardiac Biopsy Rejection: After a patient receives a heart transplant, routine biopsies are taken to determine the patient’s risk for transplant rejection. Using machine learning to model longitudinal heart biopsies, my research partners and I were able to show that we can predict future cardiac rejection events as well as risk with high fidelity.
• Vector Classification in African Countries: In order to effectively control mosquito spread, entomologists need to determine the species, sex, infectivity, and abdominal status of mosquitoes. Lack of microscopes and entolomologists is creating a major bottleneck in this workflow, thus preventing effecive mosquito control. Along with the JHU Center for Biomedical Innovation and Ghana’s Ministry of Health, I created a robust and interpretable machine learning algorithm to determine species, sex, and abdominal status of mosquitoes. The lightweight algorithm can run on mobile phones and is being tested in Ghana, Zambia, and Uganda.