A 2.82B-parameter language model trained on 20.1B tokens of pre-1931 text. Contributed the training harness, vintage evaluation benchmarks, fine-tuning, and serverless inference. Built an automated research loop that ran 100 experiments to improve the model.

An iOS workout tracker I built for my own training, but later expanded to support gym communities. Log workouts, build routines, explore an exercise library, and track strength standards and progress with interactive charts. Refined through daily use and feedback from five testers.

Hands-free computer control using a NeuroPawn BCI headset. Detects eyebrow raises and jaw clenches from IMU signals with a personalized classifier trained on over 400 samples. Built real-time signal processing and compared five model families at the Neurotech Hackathon.

Built in under 24 hours at Hack_NCState. Converts 2D product photos into AR-ready 3D models with rotate, zoom, and augmented reality preview; inspired by Amazon's view-in-room experience. Users can upload or snap a photo, enhance it with Google Gemini Nano Banana, and export an interactive GLB. Demo generation takes about 4 minutes and costs around $0.25.

A social app for discovering movies and TV shows, built with a five-person team at NCSU App Club. Includes TMDB search, watchlists, episode discussions, and local caching. Led the SwiftUI architecture, development, and code reviews.

A neural-network competition I created for Thinking Machines. Train a model to approximate the Mandelbrot set with just 30 seconds on an NVIDIA T4. Includes a Colab notebook, automated evaluation, and a live leaderboard with 217 submissions.

Designed and built the website for a commercial glass and glazing contractor through RDU Technology. Delivered in June 2026 with 100 Lighthouse scores across Performance, Accessibility, Best Practices, and SEO.

Serverless SaaS dashboard with PostgreSQL, Edge Functions, Row Level Security, Stripe subscription management, and real-time call analytics processing 100+ daily events with automated data sync. Developed admin panel with business account management, call history filtering for serving Triangle area businesses.

Deep learning model classifying breast histopathology images as benign or malignant. Trained ResNet34 on 555,000+ medical images from 279 patients achieving 88.6% accuracy; built baseline ResNet18 on 143 web-scraped images at 82% accuracy.

1st place hackathon winner. AI-powered dementia caregiver app built on the Johns Hopkins DICE framework, featuring voice-first episode logging, pattern detection to predict and prevent behavioral episodes, and a personalized Memory Book. Targets the 89% of unpaid dementia caregivers who have received zero formal training.

Led frontend development team for a mobile application that helps students forecast their GPA and receive personalized academic advice by uploading transcripts to Google's Gemini API. Developed GPA Forecaster screen with dynamic course modification and transcript upload functionality. Integrated backend+frontend for modifying courses, terms, and transcript data with SQLite.

1st place, Wright Start Accelerator. Co-founded and built responsive breast cancer awareness website featured on the news, serving 200+ Raleigh community members with donation, hero nomination, and volunteer signup capabilities. Built with React, TypeScript, and Tailwind CSS.

Featured in Harvard's CS50 Gallery. Decision-making web app with weighted pros-and-cons, user auth, archiving, real-time visualization, and recommendations. Built with Python, Flask, and PostgreSQL.

Built beverage ordering system as CSC 326 capstone with a team of 4 using MVC architecture and RESTful APIs. React frontend with Spring Boot backend featuring user auth, order management, inventory tracking, and full CRUD operations with MySQL. Created comprehensive Figma design system.

Automation tool for dynamic pricing updates across 1,000+ trading cards, saving local stores 5+ hours of manual work weekly. Built with Python, Playwright, and Pandas.

Pomodoro Technique app with analytics using Kotlin and Figma. Increased user productivity by 50% in beta testing with comprehensive time tracking features.

A betting app for NC State sports game winners with live statistics, leaderboards, and championship recognition. Made in 1 hour during SOE Coding Hackathon.