Lingyi Zhou
Portfolio Website
 All My Works

Ex. AR Engineer @ Snap Inc.
Ex. Unity Engineer @ Unity

MFA Design and Technology @ Parsons
MA Game Design & Development @ Columbia XR Developer
AR Development in Lens Studio & Effect House.
VR Development in Unity for Oculus Quest and Apple Vision Pro Game Developer
Unity Development for desktop and mobile games
Shader knowledge and development experience 3D Designer
Blender and Maya for stylized and realistic modeling and rendering  
Cinema4D for animation and procedural modelling  
 Full Stack Engineer
React front end + FastAPI Backend + Postgre database Graphic Designer
Adobe Indesign, Adobe Illustrator

CuraLoop: AI Companion for Alzheimer's Care (2025)

🏥 1-day Humans & AI Hackathon project in Seattle

Monitors cognitive health through AI conversations and gamified exercises to predict Alzheimer's care needs. Built with a Human-in-the-Loop model connecting patients, caregivers, and doctors for continuous, empathy-driven care.    

Tech Stack (Built in 6 hours):
  • Frontend: React Native (TypeScript) via Expo
  • Backend: FastAPI (Python)
  • AI/ML: Claude Sonnet Model, Custom trend analysis
  • Deployment: Vercel
  • Team: 3-person collaborative development

Links:

The Problem

    Alzheimer's symptoms often progress silently between infrequent clinical visits. Current care models rely on subjective, fragmented assessments, which means cognitive decline often goes unnoticed until it becomes severe.    

    This leads to:    
  • Delayed interventions
  • Missed opportunities for improving quality of life
  • Increased strain on caregivers
  • Lack of continuous monitoring data for doctors

The Solution

    CuraLoop introduces an AI-powered daily companion that continuously tracks, predicts, and supports a patient's cognitive and emotional well-being.    

    Through daily check-ins, gamified cognitive exercises, and natural language conversations, the platform captures subtle behavioral, linguistic, and emotional signals.    

    Our AI algorithms analyze these trends to detect potential deterioration, issuing alerts to doctors or caregivers to enable proactive and early intervention.  

Core Features

🏠 Home Screen

The main interface where patients start their day. Features a clean, senior-friendly design with large buttons and clear navigation to daily check-ins, games, and chat functions.    

Key Elements:
Through Chatbot:
  • Quick access to daily check-in
  • 24/7 chat companion entry
  • Progress tracking display
  • Upcoming reminders and notifications

Through Game:
  • Game library navigation
  • Game behavior recording
  • Cognitive game as brain exercise

💬 24/7 AI Chat Companion

    An empathetic AI assistant powered by Claude Sonnet that patients can converse with at any time. The system analyzes language patterns, emotional tone, and conversation coherence to identify confusion, depression, or behavioral changes.    

Features:
  • Natural language conversations
  • Emotional sentiment analysis
  • Memory recall prompts
  • Confusion detection and gentle redirection
  • Real-time empathetic responses
  • Conversation history logging for doctors

🎮 Gamified Cognitive Training

    AI-adaptive games designed to enhance memory, attention, and orientation while generating quantitative performance data. Games adjust difficulty based on patient performance to maintain engagement without frustration.    

Game Types:
  • Memory matching exercises
  • Pattern recognition challenges
  • Word recall games
  • Visual-spatial puzzles
  • Attention training activities

    Each game session generates performance metrics that feed into the AI analysis system for trend detection.  

🩺 Doctor Dashboard

    A comprehensive web dashboard for healthcare providers that displays patient cognitive trajectories, emotional trends, and detailed activity logs. Doctors can review AI-generated insights and make informed care decisions.    

Dashboard Features:
  • Patient overview with key metrics
  • Cognitive performance trend charts
  • Emotional health indicators
  • Alert review and approval system
  • Conversation transcript analysis
  • Treatment plan modification tools
  • Export reports for clinical records

Additional Features

🗓️ Daily Check-Ins

    Establishes a consistent cognitive baseline through simple memory and orientation questions. Patients answer questions about the date, recent events, and personal information to track day-to-day cognitive stability.    

🔔 Predictive Alert System

    AI models continuously analyze user data to detect negative trends such as declining memory performance, increased confusion, or emotional distress. When concerning patterns are identified, the system automatically triggers notifications to designated caregivers or doctors for immediate attention.  

⚕️ Human-in-the-Loop Review

    Doctors receive auto-generated risk summaries with AI-detected patterns. They can review findings, approve or modify assessments, and update treatment plans. This ensures AI insights are validated by medical professionals before any care adjustments are made.    

📊 Continuous Monitoring

    All patient interactions—from games to conversations—are logged and analyzed to create a comprehensive picture of cognitive health over time. This longitudinal data enables more accurate predictions and personalized care strategies.  

AI + Human-in-the-Loop Workflow

    Our system creates a continuous feedback loop that ensures both precision and safety:    

  1. User Interaction - Patients engage with daily check-ins, games, and conversations
  2. Data Logging - All interactions are securely captured and timestamped
  3. AI Analysis & Pattern Detection - Claude Sonnet and custom ML models identify trends
  4. Risk Alert Generation - System automatically flags concerning patterns
  5. Doctor/Caregiver Review - Human experts review AI findings and context
  6. Treatment Plan Adjustment - Approved changes are implemented
  7. Continuous Feedback Loop - Updates inform future AI predictions

        This process leverages AI for real-time monitoring and prediction, while human experts provide the critical oversight needed for safety, interpretability, and ethical decision-making.  

System Architecture

Mobile Application (React Native)

Frontend/
├── Daily Check-in Module
├── Game Engine
├── Chat Interface
├── Alert System
└── Offline Sync Handler
  • Built with Expo for cross-platform deployment
  • TypeScript for type safety
  • Offline-first architecture with sync
  • Real-time chat with Claude Sonnet API

Backend API (FastAPI)

Backend/
├── User Management
├── Data Analytics Engine
├── ML Trend Analysis
├── Alert Trigger System
└── Doctor Dashboard API
  • RESTful API with FastAPI
  • Python-based ML models for trend detection
  • Integration with Claude Sonnet for conversations
  • Automated alert generation

24-Hour Development Sprint

Technical Challenges & Solutions for building the app in 6 hours

Challenge 1: Real-time AI Analysis
Problem: Need to analyze conversations in real-time without lag
Solution: Implemented streaming responses from Claude Sonnet API with client-side caching    

Challenge 2: Trend Detection
Problem: Limited time to build complex ML models
Solution: Used statistical analysis with rolling averages and anomaly detection    

Challenge 3: Human-in-the-Loop Balance
Problem: Balancing automation with human oversight
Solution: Created tiered alert system—minor changes logged, major changes require doctor approval  

INDEX  NEXT