canvas
Team led by an EY Malaysia Cyber Security Consultant (4y) with MSc Data Science (City, U. of London), Python/FastAPI/LLMs, GCP, and Malaysian Lead Generator experience.
YouTube Video
Project Description
Problem Statement
Healthcare professionals spend 2+ hours daily on clinical documentation, leading to burnout and reduced patient interaction time. Current EMR systems present information in linear, text-heavy formats that don’t match clinical thinking patterns. Critical patient information gets buried in complex documents, delaying decision-making.
Proposed Solution
HospitalCanvas is an AI-powered clinical intelligence platform featuring:
- Interactive Canvas Interface: Drag-and-drop clinical modules replacing traditional EMR navigation
- RAG-Powered Q&A: Semantic search across patient documents with source attribution
- Automated SOAP Generation: AI clinical documentation meeting professional standards
- Visual Data Integration: Real-time charts, timelines, and analytics dashboards
Target Users & Workflow
Primary Users: Emergency physicians, specialists, nurses, clinical analysts
Core Workflow:
- Patient selection → Canvas loads with relevant clinical modules
- Drag/resize modules to match clinical priorities
- AI Q&A for instant patient insights with source citations
- Generate professional SOAP notes automatically
- Visual analytics for population health and treatment effectiveness when switched to Analyst view
Measurable Impact
- Time Savings: 85% reduction in documentation time (2+ hours → 20 minutes)
- Clinical Accuracy: AI responses cite source documents with confidence scoring
- User Experience: Intuitive canvas interface requiring minimal training
- Scalability: Microservices architecture supporting enterprise deployment
Judging Criteria Alignment
Innovation: First canvas-based clinical interface combining RAG AI with visual workflows
Feasibility: Fully functional demo with production-ready architecture
Technical Merit: Advanced RAG pipeline, vector embeddings, real-time canvas synchronization
User Benefit: Dramatically reduces administrative burden while improving clinical insights
Deployability: Cloud-native architecture (Netlify + Railway) ready for healthcare environments
Technology Stack
Frontend: React 19.1, TypeScript, @xyflow/react, Tailwind CSS 4
Backend: FastAPI, SQLite, Pydantic validation
AI/ML: OpenAI GPT-4, Sentence Transformers (BAAI/bge-large-en-v1.5), FAISS vector search
Infrastructure: Netlify (frontend), Railway (backend), SQLite (data persistence)
Security: HIPAA-compliant data handling, synthetic patient data for demos
Demo Validation
- Live deployment: hospitalcanvas.netlify.app
- Comprehensive Uncle Tan case study with complex chronic kidney disease
- Real-time AI responses with sub-2-second performance
- Professional-quality SOAP note generation validated against clinical standards
Disclaimer: All data is seeded synthetic data, no real PII or PHI was used in the use of this project
Prior Work
The main layout canvas and nodes was made a few days after the problem statement was released