Christy Li

I am an undergrad at MIT studying CS and Physics. Currently, I'm researching automated harm remediation for computer use agents as a Research Scholar at ML Alignment and Theory Scholars (MATS). I am broadly interested in building AI agents that can learn and adapt through interactions with their environment and users to effectively collaborate with humans on complex, long-horizon tasks.

In my free time, I like to read, play volleyball, and slowly but surely taste test every matcha drink in Boston. If you'd like to chat about any of the above or just want to say hi, feel free to reach out!

Christy Li

Research

Other Projects

FPGA
FPGA Lab (6.2050) Final Project

Optical FlowPGA

Implemented an FPGA-based motion tracking system for real-time detection and visualization of salient feature points on moving objects in video streams.

Report
Deep Learning
Deep Learning (6.7960) Final Project

World Modeling with Graph Neural Networks

Trained GNN-based structured world models for next-state prediction in toy environments including block pushing, n-body gravitational physics, and Atari games, achieving up to 100% accuracy for short-horizon predictions.

Report
RedAgent

RedAgent

A GPT-4o agent for context-aware red teaming of LLMs, based on the RedAgent paper. Multi-agent system with long and short-term memory capabilities built with LangChain/LangGraph.

Code
booksalotl

booksalotl

MERN stack web application for tracking and rating books and generating and sharing personalized reading statistics. Integrated Google OAuth for secure user authentication and Google Books API for book search.

Code