Projects
Personal and Research Projects
Visualizing CHI Research with Generative AI
With CHI growing from 60 papers a year to over 1,000, tracking research trends by hand is no longer feasible. This project uses Generative AI to sort 11,847 CHI abstracts into 921 topics under 26 broad categories. A streamgraph and heatmap reveal how the field has shifted over three decades, showing which topics are growing and which are fading.
Research · Generative AI, Text Embedding, Clustering, Data Visualization, NLP, HCI
Commit Agent CLI
An AI-powered git commit message generator using Claude Sonnet 4.5 / Opus 4.5 or Google Gemini and LangGraph. Generate intelligent, context-aware commit messages by simply typing `commit`.
Developer Tools · Node.js, TypeScript, Claude AI, Google Gemini, LangGraph, CLI, Git, AI Integration
Fintellection
An AI-powered financial intelligence platform that brings institutional-grade research capabilities to retail investors. The MVP features an Agentic Researcher that conducts multi-step financial analysis grounded in verified data sources like SEC filings, and a Stock Dashboard built on TradingView that visualizes research outputs alongside interactive price charts. Built on Next.js, Supabase with pgvector for RAG, and Vercel AI SDK with Mastra for agent orchestration.
FinTech · Next.js, Supabase, Vercel AI SDK, Mastra, RAG, TradingView Charts, Financial Data APIs
Token-Efficient Video Game Summarization for Multimodal LLMs
Event-driven frame selection for LLM-based video summarization using Apex Legends gameplay with the Overwolf API. By capturing screenshots only when key events occur, the method achieved 78.1% accuracy with 73% fewer tokens compared to raw video processing while improving accuracy by 14%.
Research · Python, OpenCV, GPT-5, Claude, Overwolf API, Video Processing, LLM, Multimodal AI
Scholar Capital
A web application that transforms Google Scholar profiles into interactive network visualizations. By simply replacing 'scholar.google.com' with 'scholar.capital' in any URL, users can explore research networks through dynamic graphs, featuring progressive disclosure of papers, co-author connections, and citation analysis for enhanced academic research discovery.
Web Development · Network Visualization, Data Visualization, Google Scholar API, Interactive Graphs, Citation Analysis, Web Application
Foodiverse
A pixel-art campus game where students can discover and share food resources through community fridges and mutual-aid listings. The platform features interactive 2D navigation, real-time updates, AI integration for food identification, and comprehensive accessibility design to address food insecurity while building community.
Social Impact · Next.js 15, TypeScript, PostgreSQL, Google Gemini AI, Pixel Art, Game Development
Instant Prompt Optimizer
A Chrome extension that instantly optimizes highlighted text into better AI prompts using Google's Gemini AI. The extension transforms AI conversations with improved, more effective prompts across popular platforms like Claude.ai, ChatGPT, Google Gemini, and Perplexity.
AI Tools · Chrome Extension, Google Gemini AI, JavaScript, Prompt Engineering, Browser Extension API, AI Integration
Can I Wear Shorts Now?
A weather-based web app that tells you whether it's suitable to wear shorts today based on temperature and weather conditions.
Web Application · Next.js, TypeScript, OpenWeather API, IPInfo API, Geolocation, Tailwind CSS
AI-Assisted Diary Study Platform for HCI Education
A web-based platform that streamlines HCI-focused diary studies with intuitive submission tools, progress tracking, and sentiment analysis. The system uses a fine-tuned Mistral 7B model to generate constructive, rubric-based feedback and thematic insights from student reflections, enabling scalable AI-assisted evaluation and promoting deeper engagement in large classroom settings.
Education Technology · Web Development, AI/ML, Mistral 7B, Sentiment Analysis, HCI
DigitalPage
Comprehensive speech-to-speech pipelines integrating OpenAI's Whisper model with advanced VAD techniques to create seamless audio processing workflows. The implementation enables real-time voice interactions with optimized latency and enhanced user experience through intelligent speech detection and processing.
Voice Technology · OpenAI Whisper, VAD, Speech Processing, Real-time Audio, Pipeline Architecture
Public Sentiment on Generative AI in CS Education
A research project analyzing public sentiment towards Generative AI in CS education using BERT-based models and Large Language Models (LLMs). The study collected and filtered Reddit comments to identify key concerns and benefits of Generative AI tools in educational settings, revealing positive sentiment towards AI for learning while highlighting challenges such as academic dishonesty and over-reliance on AI.
Research · BERT, Large Language Models, Social Media Analysis, Sentiment Analysis, Data Mining