In summer 2023, I was a research fellow in Program for Research in Science and Engineering at Harvard University researching formal verification methods for large language models (LLMs) under Professor Nada Amin as part of the Programming Languages Lab. During my internship, I refined Parsel, a decomposition framework for developing Python programs more accurately, to work with Coq, an interactive theorem prover. I developed a ChatGPT plugin to verify Coq theorems and combined this with a novel tree-structured prompting technique that I created to improve LLM reasoning skills directed toward theorem-proving. I also fine-tuned an open-source LLM transformer model using PyTorch to improve Coq language comprehension.

Worked with students from the University of Pennsylvania to perform data analysis on and create models predicting Delta Airlines stock price as a function of various revenue, cost, and customer satisfaction factors. Invited to compete at the Citadel Data Open Championship in Fall 2023.

Worked with fellow students from Harvard University, Northeastern University, and Worcester Polytechnic Institute to develop Oasis, a decentralized social media platform for users to freely converse with each other through posts and comments while profiting from popular self-curated content. Built platform using the Staking DAO protocol on top of the Moonbase parachain through Polkadot, Crust, and React. Won 1st place in the Moonbeam project track, runner-up in Grand Prize Polkadot track.