Full-stack apps, ML research, and web development work
Voice-first food tracking iOS app built in 3 hours at the Georgia Tech x Anthropic hackathon (featured by NBC News). Gamified fitness tracking with health tokens, streaks, and consistency rewards. Secured finalist placement among 20+ competing student teams.
Interactive design token editor that generates 7-category token sets from natural language prompts. Uses the Figma Variables API and a fine-tuned small language model to automate token generation covering color, typography, spacing, and more for various UIs.
Unified Georgia Tech app consolidating 15+ scattered campus websites and tools. Features campus bus tracking, dining menus, meal pass integration, study room booking, and real-time crowding data for dining halls, libraries, and the campus recreation center.
Real-time multiplayer chess platform with WebSocket-based matchmaking, Stockfish AI (web worker), voice chat, lobby system, friend requests, and JWT auth backed by PostgreSQL via Prisma.
Offline-first PWA for journalists and war correspondents. AI coach powered by Claude API with full health context (nutrition, sleep, schedule). Google Calendar sync, IndexedDB local storage, installable as a native app.
Student-made learning hub teaching HTML, CSS, and JavaScript to high school students. Features structured curriculum by unit, peer tutor booking, session scheduling, a student dashboard, and supplemental resources. FBLA 2026 — 2nd at NC States, qualifying for Nationals.
Job board platform built for 3,600+ high school students and their counselors. Features employer posting, student applications, input authentication, and backend approval workflows. FBLA 2025 National Leadership Conference Finalist.
Cross-retailer product matching pipeline for Sephora and Ulta. Scrapes product listings, collapses shade variants, matches equivalent SKUs across retailers using sentence embeddings + Union-Find clustering (cosine similarity > 0.88), and surfaces price gaps in a Next.js dashboard with Recharts.
AI-powered document template filler. Upload a source document, describe what you need in plain language, and the app generates a completed template. Built with Next.js and the Claude/OpenAI API.
Graduate-level computer vision research at Georgia Tech. Developed CNN-ViT hybrid models for domain adaptation, exploring cross-domain visual feature transfer with PyTorch and TensorFlow.
Full-stack web app for identifying Georgia Tech campus landmarks from uploaded photos using a ResNet-18 CNN. React front-end, Python backend with PyTorch model inference, MongoDB storage.
Research fellowship at Algoverse. Conducted literature reviews and benchmark testing on legal text and sentiment analysis models using Hugging Face transformers, BERT, and GPT-based architectures.
Location-based cafe discovery app fetching real-time data via Maps API. Clean, accessible mobile-first UI built with vanilla JS and CSS.
Mobile-first web app replicating the CVS Pharmacy experience. Features a splash screen, responsive grid layout, sticky footer, and adaptive header styling across mobile and desktop.
Modern responsive landing page for a cloud solutions company. Showcases services, products, and client benefits. Includes homepage, services overview, and contact sections.
Full-screen landing page with a background video and smooth overlay effect. Toggleable navigation menu, bold typography, and responsive design for desktop and mobile.