Course 6-4 and 15-2 Undergraduate Class of 2025
Massachusetts Institute of Technology

I was born in Salt Lake City, Utah, but spent most of my life growing up in Shanghai, China. I attended high school in Pottstown, Pennsylvania. Currently, I'm based in Boston as a college student.

Hobbies: DJ, Snowboarding, Traveling

Coursework (GPA: 4.8/5.0)

Fall 2024

  • 6.7960(G): Deep Learning

  • 6.7900(G): Graduate Machine Learning

  • 6.2400: Quantum Systems Engineering

  • 6.C571: Optimization Methods

Spring 2024

  • 6.1220: Design & Analysis of Algorithms

  • 6.8300(G): Advances in Computer Vision

  • 6.S041: Algorithmic Solutions to Human Problems

  • 15.075: Statistical Thinking and Data Analysis

  • HAA.2002: Decisions Big & Small (Harvard cross-register)

Fall 2023

  • 6.4110: Representation, Inference & Reason in AI

  • 6.8611(G): Quantitative Methods for NLP

  • 18.03: Differential Equations

  • 5.111: Principles of Chemical Science

Spring 2023

  • 6.4200: Robotics, Science & Systems I

  • 6.3100: Dynamical System Modeling and Control Design

  • 14.13: Psychology & Economics

  • 7.014: Introductory Biology

  • 21G.111: Chinese Calligraphy

Fall 2022

  • 6.3900: Introduction to Machine Learning

  • 18.600: Probability & Random Variables

  • 18.06: Linear Algebra

  • 15.401: Managerial Finance

  • 15.501: Corporate Financial Accounting

  • 21M.710: Script Analysis

Spring 2022

  • 6.006: Introduction to Algorithms

  • 6.009: Fundamentals of Programming

  • 8.02: Physics II

  • 21W.015: Writing and Rhetoric on Sports

Fall 2022

  • 6.0001: Intro to CS Programming in Python

  • 6.042: Mathematics for Computer Science

  • 18.01/18.02: Calculus

  • 14.01: Principles of Microeconomics

  • 8.01: Physics I

  • 9.01: Introduction to Neuroscience

Projects

Curator AI

Mantis AI

Mantis is pioneering the next wave in scalable AI, developing a state-of-the-art conversational "visual database." This tool integrates LLMs to sift through vast datasets, offering leaders a panoramic view of their organizations, from personnel and teams to comprehensive project overviews. Our system not only lays out essential context but also delivers proactive insights, streamlining the supervision and management process.

To learn more:

Website

Github Repo

Mantis Memo

Pitchdeck

BeaverNav

BeaverNav is a web and mobile application designed to guide users through complex indoor environments using real-time navigation. It transforms floorplans into detailed maps and uses path-finding algorithms to chart the shortest routes, initially focusing on MIT before expanding to other venues.

To learn more:

Github Repo

Pitchdeck

HalluciBERT

Football Snap Detection

We have developed HalluciBERT, a BERT-based model designed to predict the likelihood of hallucinations—instances of false information relayed with high confidence—in LLMs before inference. While existing mitigation methods often match the computational size of LLMs themselves, HalluciBERT offers a more efficient approach. In tests, it surpassed baseline models in predicting confidence scores and demonstrated a strong correlation with ground truth, although its precision in scoring accuracy requires further enhancement.

To learn more:

Final Research Paper

Self Driving Toy Car

StockSwift

I had the opportunity to take Robotics: Science & Systems in the Spring of 2023. Throughout the semester, we dived into labs that spanned a wide range of topics including kinematics/dynamics, controls, sensing and computer vision, localization, path planning, and motion planning, all using Python & ROS.

To learn more:

Lab Portfolio

This project focuses on improving how we detect the exact moment of the snap in football games using video footage from the MIT Football Team. We developed a real-time processing model that is optimized for understanding spatial information in the game, and we compared it to Hudl's current method, which is based on image segmentation. Our model proved to be just as accurate as Hudl’s but delivered better quality predictions, making it a more cost-effective and efficient tool for football analysis.

To learn more:
Final Research Paper

Final Presentation

StockSwift is a platform designed to connect retail vendors with on-demand gig workers for merchandising tasks, like stocking and organizing products on store shelves. We focus on addressing inefficiencies in retail merchandising caused by labor shortages, high turnover, and unreliable service from existing third-party merchandisers. Leveraging geolocation technology, StockSwift optimizes task assignment by grouping jobs in close proximity, ensuring that gig workers can complete multiple tasks in one shift, reducing travel time, and increasing job satisfaction. By providing real-time tracking, instant payments, and data analytics on product stock levels, StockSwift offers a cost-effective, scalable solution for vendors while providing flexible work opportunities for gig workers.

To learn more:
Pitch Deck

Landing Page

Robotic Arm

I built a 3D printed robotic arm.This system operates using servo motors and fishing strings to facilitate movement. Additionally, it uses computer vision to accurately detect hand motions, allowing it to skillfully mimic finger movements.

Curator.ai is an AI-powered platform designed to make furniture shopping smarter and easier. It provides context-aware recommendations based on your space, style, and budget, using AR tools to help users visualize furniture in their rooms before buying. The platform integrates with multiple vendors to offer real-time suggestions, ensuring users can make informed decisions quickly and confidently.

To learn more:

Website

Pitchdeck

more coming soon!