
Which LLM Fixes It Best? A Comparative Study of Automated Program Repair
Comparative benchmark study evaluating leading LLMs on automated program repair across diverse bug types.
A collection of AI systems, web applications, and engineering solutions I've built for clients and personal exploration.

Comparative benchmark study evaluating leading LLMs on automated program repair across diverse bug types.

Professional marketing website with admin CMS for a construction company, featuring a 3D animated logo and full content management system.
Evaluates differential privacy as a defense against membership inference attacks in federated learning, revealing counterintuitive results where DP increased attack success on overfit models.
Multimodal AI app comparing LLaVA 1.5 vs 1.6 for generating personalized recipes from fridge/pantry photos based on dietary goals.
A full-stack Twitter-like social media platform with web, API, and mobile apps supporting real-time notifications, multi-language i18n, and comprehensive security.

A production-deployed full-stack portfolio platform built with Next.js, featuring 76 React components, 32 RESTful API endpoints, and a custom CMS. Includes a Three.js-powered 3D animated background, real-time analytics with 13 custom events, and comprehensive test automation with 2,962 tests achieving 77% code coverage.
A Linux kernel module implementing a bar simulation with waiters, customers, tables, and waiting queues using kthreads and system calls.
A custom Unix shell implementation in C featuring command parsing, I/O redirection, piping, background processes, and built-in commands.
A secure news aggregation site powered by Hacker News API with Auth0 authentication and automated hourly article updates.
You've seen all 9 projects