Shu Yang

Abstain Mask Retain Core: Time Series Prediction

San Diego, CA | top 5% of accepted papers

Designed an interpretability framework to visualize manifold geometry and Stochastic Approximation (as seen in the diagrams), verifying how the Embedding Similarity Penalty prevents representation collapse. Conducted rigorous ablation studies across diverse datasets to validate the robustness of Adaptive Masking Loss.

• Read Paper
• View Details

SomniBot: AI-Powered Personalized Sleep Assistant

An AI-driven sleep assistant providing personalized, science-backed guidance using FHIR and RAG. It features a self-assessment quiz and intelligent analysis based on EHR data to improve sleep quality while ensuring privacy.

Learn more

CLT Legal AI Assistant: Intelligent Chatbot & Multilingual Platform

An AI-powered legal support tool for Community Land Trusts (CLTs). It uses machine learning to parse legal datasets and was iteratively enhanced across three versions to include a multilingual chatbot, user accounts, and an interactive map.

Learn more

Dauyan AI Writing Assistant

This project is a website which trying to combinate AI and English writing in order to help students study and exercise it deeper and easier. As AI could help them in conception, writing, scoring and other personalized questions

Learn more

Houston's Healthcare Divide: Mapping Access and Inequality

This study uses GIS network analysis to examine healthcare accessibility in Houston. By analyzing road networks and travel times, it reveals disparities in resource distribution and its impact on residents.

Learn more

Economic Toll of Flooding: Florida Case Study

This project uses GIS and machine learning to predict flood risk and economic losses in Florida. It models vulnerability and economic impact by integrating physical, demographic, and socioeconomic data.

Learn more

Facial Pixel Mosaic: Real-time Generative Art with Eye Tracking

An interactive artwork using ml5.js FaceMesh to create dynamic mosaic portraits from live video. It features eye-triggered regeneration and demonstrates creative coding with computer vision.

Learn more