Designed and implemented a custom Model Context Protocol (MCP) server integrated with Qdrant Cloud for semantic recipe search and generation. Built a Python-based crawler to collect and parse structured recipe data from Wikibooks, stored it in a vector database with embeddings for fast retrieval, and exposed structured MCP tools for seamless interaction with Claude. Delivered a scalable, cloud-ready architecture capable of full-text and semantic filtering, supporting multilingual expansion and further integration with LLM-driven recipe creation workflows.

EatMeFirst is a smart refrigerator management app that empowers users to take control of what’s inside their fridge. It tracks expiration dates, recommends what to eat first, suggests recipes based on available ingredients, and helps you build smarter shopping habits — all from your phone.

Part Select Agent is an AI-powered chatbot that helps users select the right parts for their refrigerator. It uses a combination of fuction calls and polymorphism to handle different user queries and provide accurate recommendations, based on Embeddings and RAG.

Render Fusion is a real-time remote rendering pipeline that allows users to render 3D objects in a web browser. It offloads rendering to a nearby machine and streams the results to a web browser. I was part of the team that developed the renderer, including the implementation of the rendering pipeline, the integration of the renderer with the web application, and the optimization of the renderer for real-time performance.

This is a React + TypeScript + Vite web application that visualizes weather balloon data from WindBorne Systems and integrates with Open-Meteo weather data to provide wind field analysis.

Rhythm Quest is a web-based rhythm game that allows users to play music and dance to the rhythm of the music. I developed the game using WebGL and Three.js in my free time.

PetPal is a web application that builds up a friendly community for pets. It allows users to help their pets find friends based on AI scoring and distance. It uses a combination of React and Django to create a full-stack web application.

In the era of COVID-19, although with vaccines, the primary line of defense still lies upon face masks. In many countries, we have seen that there are manual people checking whether or not people are wearing masks properly. Therefore, the idea of creating a mask detector sparks our interest. In this project, we aim to apply computational photography, computer vision, and machine learning to create a face mask detector. During this process, we hope to apply what we have learned during the class and challenge ourselves with essential techniques in Computer Vision that can potentially improve our world.