Austin, Texas, United States
I'm a student at the University of Texas at Austin pursuing a major in computer science and a Bridging Disciplines Program certificate in Human Rights and Social Justice. I've always been passionate about technology, as well as working with and helping others. I'm seeking work involving computer science that also has a social impact.
- Supporting a nonprofit organization dedicated to empowering Asian and immigrant survivors of domestic violence, sexual violence, and human trafficking by strengthening data systems and grant reporting efforts. - Assisting with collecting, cleaning, and analyzing program data to support grant compliance and organizational decision-making. - Pulling reports from internal databases, researching external data sources to strengthen grant narratives, and collaborating cross-functionally to understand data needs across programs. - Maintaining strict confidentiality while working with sensitive client data and contributing to administrative and reporting workflows that support long-term sustainability of services.
Responsible for integrating generative AI capabilities into the core supply chain platform to provide secure, natural-language data access for enterprise planners. Key Contributions: - Designing the core architectural components for a secure, multi-tenant LLM-powered Text-to-SQL system. This includes implementing Approved-Views Registries and query-validation logic to enforce tenant isolation and ensure explainable SQL generation across massive supply chain datasets. - Spearheading the exploration and architecture of advanced retrieval methods (RAG and Temporal-based architectures) to enhance domain-specific grounding and contextual relevance. Collaborating with cross-functional teams (DevOps, UI, and Data Governance) to define the path for the next-generation, scalable AI assistant.
As a Research Assistant in a biomedical engineering lab, I applied advanced deep learning techniques to tackle critical challenges in medical imaging. • Trained a 3D convolutional autoencoder using MATLAB’s Deep Learning library to efficiently analyze and process lung data, applied mean absolute error loss to accurately reconstruct and identify tumors in test scans • Developed a lung segmentation model with UNet architecture for classifying lung structures in CT images using MATLAB’s Deep Learning library, created custom pixel-wise labeling and utilized class weighting for improved segmentation accuracy • Designed an algorithm using connected component analysis to segment liver vasculature, aimed at creating a predictive model for assessing the success of embolization noninvasive medical procedures (better alternative to open surgery) • Engineered a deep learning pipeline, TriSwinUNETR, for lung lobe segmentation, achieving state-of-the-art DICE accuracy of 93.72% • Optimized Python code to calculate breathing-induced volume changes from lobe segmentations, reducing code size by 20%
I built an AI-powered summarization tool to help cardiologists quickly understand patient history before interpreting echocardiograms. This project leverages large language models (LLMs) to generate clinically relevant summaries from unstructured input data, reducing the manual chart review burden and improving provider efficiency. My responsibilities encompassed: -Designing and implementing a modular agent-based summarization pipeline using C# and Epic’s internal AI orchestration platform -Engineering prompt templates and dynamic token injection logic to allow flexible summarization across specialties and study types -Collaborating cross-functionally with clinical informatics experts, AI engineers, and platform engineering teams to design for seamless integration with Epic's evolving systems and ongoing initiatives