My current interests are in machine learning, artificial intelligence, and agentic systems, particularly their applications in computational biology, trust and safety, and beyond.
Advisors: Prof. Eliot Moss, Prof. Philip Thomas
We investigate how errors from floating-point arithmetic, and how more accurate methods such as compensated summation, affect the training of deep neural networks.
Advisor: Prof. James Allan
This Search Engine Honors project explores how different manipulation techniques influence the downstream product recommendation behavior of retrieval-augmented generation (RAG) systems. We also experiment with various defense mechanisms to mitigate such manipulations.
Advisor: Rohan Pandey
We implemented an NLP pipeline for medical text de-identification with several surrogacy techniques on the i2b2 dataset, and evaluated them using privacy metrics, downstream task performance, and adversarial GPT-5 benchmarking.
Advisors: Prof. Hung Le, An La
We researched heuristics and implemented a fast tree decomposition algorithm in Python, achieving significant speedups over NetworkX library on benchmark road network datasets.