My Projects

A curated selection of applications and systems I’ve built — spanning full-stack web apps, cloud infrastructure, and machine learning.

Unit Vote Tracker

Full-Stack

A Streamlit + Python application that enables weighted HOA voting with real-time dashboards, persistence, and beneficial interest calculations.

PythonStreamlitPandasPlotly

Marketplace App

Full-Stack

Full-stack marketplace with Next.js, Prisma, MySQL, and TypeScript. Features authentication, admin dashboard, and role-based access control.

Next.jsTypeScriptPrismaMySQLJWT

Vision-Language Model for SuperTuxKart

Machine Learning

Fine-tuned a VLM (vision-language model) to answer visual questions about gameplay scenes. Built a data pipeline to generate QA pairs and improve model performance.

PyTorchPythonDeep LearningTransformers

Reasoning-Based Unit Conversion with LLMs

Machine Learning

Explored reasoning with SmolLM2 for unit conversion tasks. Used in-context learning and LoRA fine-tuning to achieve 50% accuracy and 85% answer rate.

SmolLM2PyTorchLoRARFT