Education
Bachelor of Science in Mathematics-Computer Science
Brown University, Providence, RI 2021 - 2025
Computer Science Department Honors, Divison 1 Track and Field/Cross Country Athlete
House Grads + Merlin
Brown Cross Country / Track and Field
Projects
RegiViz: A Tool for Generating and Visualizing Cancer Regimens
For a master's course, CSCI 2370: Interdisciplinary Scientific Visualization, we presented a tool to generate cancer regimen visualization and analysis leveraging large language models. We completed this project in collaboration with Dr. Jeremy Warner, Dr. Sanjay Mishra, and Dr. Sandeep Jain of Brown's Warren Alpert Medical School. A preliminary course paper is available here.
ChemoExperts Baseline time to generate a visualization.
RegiViz showed significantly improved time to generate a regimen visualization.
Our tool demonstrated greater clinical accuracy across criteria that patients were shown to value in previous study polls.
Figure: Comparison of our tool's performance relative to a ChemoExperts baseline. Top row shows comparison of time to generate a regimen visualization with clinically relevant 5 and 10 minute thresholds to represent average time to treat a patient. Results showed clear improvement in our method over competitors. CaDance: create your own running playlists tailored to your cadence and perceived effort! Contact me to be added to our permitted users list, as the Spotify API is rate limited. I also generated a companion app in the native Garmin language, MonkeyC, to allow users to see what biometric data to input into CaDance on the web application.
CaDance app that can be accessed after logging in! After entering your desired cadence and perceived effort level, the app generates a custom playlist on a webplayer that you may import into your Spotify account.
Master's in AI and Computational Drug Discovery and Development (AICD3)
University of California San Francisco Expected 2025 - 2026
Projects
- Nipah Binder Competition
The Nipah virus emerged recently and has exhibited high mortality and pandemic potential. Our class is participating in a competition to identify de novo molecule candidates that bind to and inhibit the Nipah virus. We are applying several frameworks, including ProtRL and Protein Hunter to generate potential amino acid sequences. Results to come soon! Agentic Pharmacovigilance
We are developing an agentic approach to autonomously process adverse patient event details and output regulatory reports for postmarket drug surveillance. We begin by applying an NLP model to parse and sort unstructured text for relevant adverse events. Then, we utilize a RAG-based causality identification model leveraging the PubMed API to produce structured ICSR E2B(R3) and E2C(R2) reports according to FDA guideline documents. We hypothesize an agentic PV system can streamline cumbersome and time sensitive case study reporting pipelines while maintaining auditability and human levels of regulatory compliance.
Traditional Manual Workflow
→Streamlined &
Automated
Agentic AI-Powered Workflow
Figure: Comparison of traditional manual pharmacovigilance reporting (left) versus our proposed agentic AI system (right). The agentic approach automates adverse event parsing, causality assessment, and regulatory report generation, significantly reducing processing time while maintaining compliance standards.
Work Experience
Graduate Student Researcher
Hong Lab at UCSF, San Francisco, CA June 2025-Present
I am a graduate student researcher under Professor Julian Hong of the Department of Radiation Oncology at UCSF. The Hong Lab is focused on machine learning and individualized clinical care. I am leading computer vision efforts within the lab and supporting coworkers with less technical backgrounds. My current focuses on using digital pathology and clinical data for outcome and subtype prediction, which I talk more about in Research!
Undergraduate Teaching Assistant
Brown University, Providence, RI January 2025-May 2025
I was an undergraduate teaching assistant for CSCI1430: Computer Vision. I held weekly office hours on topics ranging from stereo vision to convolutional neural networks. I mentored various final project groups, one of which made a violent content detector that can be seen below!
Group Photo!
Student Project: Violent Content Detector
Detection (Artificial Intelligence) Intern
FLOX, Stockholm, SE June 2023-August 2023
At FLOX, I worked on long-ranged multimodal detection algorithms to aid their efforts in autonomous drone herding in a project with the city of Stockholm to herd canadian geese away from beaches. I made significant data contributions by implementing a stereo vision image alignment pipeline and improved mAP on combined RGB-infrared image representations.