Project Portfolio

Full-stack enterprise attendance management system with Vue 3 + TypeScript frontend and Node.js + Fastify backend. Features include JWT authentication, RBAC, RESTful API design, Prisma ORM, and production deployment. Published with company authorization.

Cross-team collaboration combining MediaPipe pose recognition with agricultural data modeling for high-accuracy pen grip detection and recommendation system.

Real-time posture detection app using OpenCV and CUDA acceleration, meeting commercial-grade frame rate standards with low latency.

Interactive snake game controlled via hand gestures using MediaPipe and OpenCV, with adaptive brightness correction and real-time fingertip tracking.

Emotion classification model based on VGG19 and ResNet18, with image preprocessing, training, visualization, and confusion matrix analysis.

Systematic comparison and evaluation of classical image segmentation algorithms including Otsu, K-Means, Active Contours, Watershed, and graph-based methods.

Using word frequency vectors and PCA dimensionality reduction instead of raw text input, combined with Random Forest and SVM models for high-accuracy email recognition while preserving privacy.

End-to-end data science pipeline for public health survey data analysis, including data cleaning, modeling, exploratory analysis, and visualization.

Interactive dashboard visualizing COVID-19 outbreak trends and spatial distribution across the United States using Dash and Plotly.

Hands-on implementation and comparison of VAE and GAN generative models on Fashion-MNIST dataset, evaluating with FID scores.

Process scheduling & task graph modeling based on research paper implementation, balancing energy consumption and task latency.

WeChat Mini Program for randomized restaurant recommendations with navigation features, including merchant-side management system.