I am currently pursuing a B.S. and MEng in EECS, concentrating in Applied Physics and Hardware for AI. I enjoy working on algorithms for robots to learn more efficiently from human demonstrations and spending my weekends optimizing inference on GPUs.
Agent One is a Programming-by-Demonstration system built with the Agentic Goal Stack framework to automate desktop applications at MIT InfoLab, with Henry Lieberman, David Canfield Smith, and Honjar Xing.
Fine-tuning Qwen3-VL for robotic navigation and OpenVLA for arm manipulation at Harvard's AI and Robotics Lab, with a focus on how world models can help robots generalize from human demonstrations.
Joined LexisNexis Risk Solutions as an intern through CodeDay Labs and worked with David Zapata on reducing codebase inefficiencies by integrating SLOCCount into HPCC Systems using ECL.
Researched in high school under Dena J. Clink at Cornell's K. Lisa Yang Center for Conservation Bioacoustics, generating 1000+ spectrograms and contributed to using CNNs to classify gibbon vocalizations from wildlife field recordings.
Led the mmWave sensor team at Carnegie Mellon to build a multi-modal system with IR thermal, mmwave, and depth sensing to push OpenFace's real-world detection accuracy from 70.3% to 99.2%. This was my first exposure to research through Pennsylvania's Governor School.
