Projects

AI & Machine Learning

Comprehensive literature review, explored limitations, proposed hybrid retrieval approach combining CDE and graph RAG.
Preliminary results show graph RAG (2024) outperforms CDE (2024) and RAG (2020) by 30% on Rouge-L F1 score.

Ran audit of X's "For You" feed, analyzing 201k tweets across nine countries, working with two PhD students.
Observed 73.1% political right-leaning bias, 29.9% US-centric bias & amplification of users @alexjones, @elonmusk.

An emotionally intelligent multi-agent conversation tool used by educators to empower students to stay relentlessly curious.
Worked with students from UCBerkeley, UWashington, CUBoulder; leveraged Hume.ai, ReactJS, Supabase.

Predicted house sale prices given a dataset of house sale data.
Selected relevant features, cleaned dataset, tested KNN, Lasso and Ridge linear regression to compare RMSE scores.

Predicted the crop yield from land based on soil and yield data from John Deere.
Trained a Decision Tree Classifier ML Model and built a frontend visualization in ReactJS and CSS.

Web Development

Developed a Streamlit web app fetching stock data from Yahoo Finance API and visualizing using Plotly.
Implemented sentiment analysis of news headlines using NLTK, introducing financial data analysis and trading concepts.

An open-source, modular toolkit that transforms images into interactive, map-like components.
Developed using TypeScript and React, with Tailwind CSS.

Smart Wheelchair-accessible Navigation system for public transport commuters in Chicago.
Uses the CTA Transit, Customer Alerts, and Google Maps APIs to calculate most efficient route.

Mobile App Development

A fresh perspective to group-bill splitting and personal budgeting for young adults, powered by Splitwise.
An Android/iOS app built using Flutter, Android emulator and Material UI, with Google Sheets backend API.

A marketplace for students to give away and collect reusable items for free, built using Flutter.

Hardware & IoT

Program to sync LED lights to song playing on Spotify, enhancing user immersion; successfully tested with 30+ songs.
Written in Go on Raspberry Pi; uses Spotify API, Bluetooth, and a custom album-art based coloring algorithm.

Algorithms & Computer Graphics

C++ Visualization that attempts to compare Pokémon strength based on their Type.
Uses Force-Based Graph Visualization and Graph Coloring to investigate Pokémon Type advantages.

Developed a backtracking algorithm to calculate the Chromatic Number of a graph.
Helped students visualize and learn Graph Coloring concepts in CS 173 - Discrete Structures as a Course Assistant.