Production-grade Model Context Protocol (MCP) server and client implementation for document management workflows. Built with Python async/await patterns, exposing tools, resources, and prompts via JSON-RPC over STDIO/HTTP transports. Implements root permissions with async cleanup, progress notifications, sampling callbacks, and comprehensive tracing. Designed with guardrails and auditability for safe AI agent automation.

Full-featured Q&A platform built with modern Next.js architecture. Implements Server-Side Rendering (SSR) and Incremental Static Regeneration (ISR) with server actions and API routes. Features include NextAuth OAuth authentication, Markdown-based Q&A with voting system, server-side search with AI-powered recommendations, and AI-assisted answer generation. Achieved perfect Lighthouse score of 100 across all metrics.

Cross-platform 2D game engine developed as part of CPSC 427 course project. Built from scratch using C++, OpenGL for rendering, and Box2D for physics simulation. Features modular architecture with AI pathfinding algorithms, collision detection, interactive UI system, and particle effects. Optimized gameplay through iterative user testing, QA feedback, and performance profiling.

Statistical machine learning model for predicting Taiwan real estate prices developed for STAT 306. Implemented comprehensive feature engineering using tidyverse, exploratory data analysis, and regression modeling. Created interactive visualizations with ggplot2 to communicate insights effectively. Achieved 81.42% adjusted R² through careful feature selection and model validation.

Comprehensive B2B SaaS analytics platform providing restaurant owners with actionable insights. Features include real-time "Dine-In" analytics, customer behavior tracking, revenue forecasting using ML models, competitive analysis, AI-enhanced review management, dynamic menu optimization, and integrated coupon marketplace. Built to scale for 100+ restaurant partners.

Social dining platform with 25K+ active users connecting people through food experiences. Features AI-powered restaurant recommendations using collaborative filtering, gamified map-based discovery, location-based scavenger hunts with rewards, loyalty points system (Dyne Bucks), social meetup coordination, and exclusive deals marketplace. Implemented real-time geospatial queries and push notifications for enhanced user engagement.

Decentralized application built on Ethereum blockchain using Solidity smart contracts. Users can send messages that are permanently stored on-chain, with automatic reward distribution (0.001 ETH) to participants. Features include wallet integration using MetaMask, gas-optimized contract deployment, and event-driven architecture for real-time updates. Demonstrates understanding of Web3 fundamentals, smart contract security, and blockchain interaction patterns.

Comprehensive management platform serving 50+ NGOs, 150+ volunteers, and 20+ organizational founders. Features include volunteer scheduling and coordination, event management system, expense tracking with budget analytics, reward and recognition system, and member database management. Built with Java using MVC architecture, implements data persistence with JSON serialization, features custom GUI using Swing API with integrated audio feedback for enhanced UX.

Machine learning-powered music recommendation system analyzing audio features to suggest compatible songs for mashups. Implemented classification algorithms based on tempo (BPM), key, energy levels, and release dates. Model trained and validated on the Million Song Dataset with feature extraction using librosa. Achieved 75%+ accuracy in predicting song compatibility for mashup creation using ensemble methods.

Community-powered distributed computing platform connecting researchers with idle computational resources. Built during 24-hour hackathon using React-Bootstrap and TypeScript. Implemented real-time resource allocation system, asynchronous task queue management with Firebase Functions, and live dashboard showing computational contributions. Features include user authentication, Firestore real-time database for tracking compute jobs, and responsive design.

Statistical analysis and machine learning project predicting tennis match duration with 70% accuracy. Applied K-means clustering, classification algorithms, hypothesis testing, and data preprocessing on 5000+ best-of-3 match observations. Utilized ggpairs for exploratory data analysis, feature engineering to identify key performance indicators, and cross-validation for model selection. Implemented data visualization using ggplot2 to communicate insights effectively.

Enterprise-grade hospital management system with menu-driven interface built in C++. Features automated patient assignment based on ailments and specialist availability, attendance tracking for medical staff, floor and wing management, appointment scheduling system, and local file-based database using I/O streams. Implements efficient data structures for patient records, doctor schedules, and resource allocation algorithms.