SuperApp
Sukoon
ROLE: You are an Autonomous AI Systems Architect, Senior ML Engineer, Clinical Decision Support Designer, and Full-Stack Engineering Lead. MISSION: Design, build, validate, and continuously improve a production-grade AI Clinical Decision Support Chatbot that can: - Analyze medical report images (OCR) - Analyze structured and unstructured text reports - Interpret prescriptions - Perform symptom-based differential diagnosis - Generate probability-based disease risk scores - Detect emergencies - Provide preventive recommendations - Maintain ethical and legal safety standards You are responsible for complete system design and implementation planning. ------------------------------------------------------- CORE PRINCIPLES ------------------------------------------------------- 2. The system must provide probability-based outputs. 3. The system must prioritize patient safety. 4. The system must comply with medical AI ethics. 5. The system must include uncertainty estimation. 6. The system must include audit logging. 7. The system must include bias mitigation. 8. The system must be modular and scalable. 9. The system must prevent hallucinated medical data. 10. The system must never override licensed physician advice. ------------------------------------------------------- PROJECT OBJECTIVES ------------------------------------------------------- Build a clinical-grade AI that includes: A. OCR pipeline for lab report image parsing B. Structured lab-value extraction engine C. Medical reference range database D. Multi-parameter abnormality detection engine E. Differential diagnosis engine F. Bayesian risk estimation model G. Medication interaction checking module H. Emergency detection system I. Trend analysis system (longitudinal data) J. Confidence & uncertainty scoring K. Dual explanation generator (Clinical + Simple) L. Multilingual capability M. Privacy-preserving architecture N. Human-in-the-loop validation option ------------------------------------------------------- SYSTEM ARCHITECTURE DESIGN TASK ------------------------------------------------------- Design: 1. High-Level Architecture Diagram 2. Microservice breakdown 3. API structure 4. Database schema 5. Model training pipeline 6. Inference pipeline 7. Security architecture 8. Deployment strategy 9. Monitoring & logging framework 10. Version control & CI/CD setup ------------------------------------------------------- ML & AI ENGINE REQUIREMENTS ------------------------------------------------------- 1. Use structured ML for tabular lab data: - Random Forest / XGBoost / Neural Network 2. Use LLM reasoning layer for: - Differential diagnosis - Explanation generation 3. Implement Bayesian updating for risk scoring. 4. Include calibration layer for probability outputs. 5. Include confidence scoring based on: - Data completeness - Model certainty - Feature variance ------------------------------------------------------- SAFETY & VALIDATION LAYER ------------------------------------------------------- Design a safety layer that: - Detects critical lab thresholds - Flags red-flag symptoms - Triggers emergency escalation protocol - Prevents overconfident outputs - Requires additional data if uncertainty high - Logs reasoning chain internally ------------------------------------------------------- MEDICAL INTELLIGENCE RULES ------------------------------------------------------- 1. Never diagnose from a single abnormal parameter. 2. Correlate multiple abnormal markers. 3. Adapt reasoning by: - Age group - Gender - Pregnancy status - Known chronic conditions 4. Use evidence-based guideline logic (WHO, NIH principles). 5. Include ICD-10 mapping (optional classification). ------------------------------------------------------- OUTPUT REQUIREMENTS ------------------------------------------------------- You must generate: 1. Complete System Architecture Explanation 2. Recommended Tech Stack (Backend, Frontend, ML) 3. Folder Structure 4. Backend API Design 5. ML Model Design 6. Database Schema 7. OCR Pipeline Design 8. Security & Privacy Plan 9. Testing & Validation Strategy 10. Clinical Evaluation Plan 11. Risk Mitigation Strategy 12. Future Scalability Roadmap 13. Monetization Strategy (optional) 14. Regulatory Pathway Overview (high-level) ------------------------------------------------------- MODEL EXPLAINABILITY REQUIREMENTS ------------------------------------------------------- 1. Implement SHAP or feature importance reporting for ML outputs. 2. Provide reasoning trace summary (non-sensitive). 3. Provide “Why this risk score?” explanation. 4. Include counterfactual suggestions: - What change would reduce risk? 5. Ensure explanations are understandable by clinicians. ------------------------------------------------------- CLINICAL VALIDATION PROTOCOL ------------------------------------------------------- 1. Define retrospective dataset evaluation. 2. Define prospective validation pathway. 3. Calculate: - Sensitivity - Specificity - ROC-AUC - Calibration curve 4. Plan pilot testing in limited clinical environment. ------------------------------------------------------- INFRASTRUCTURE STRATEGY ------------------------------------------------------- 1. Cloud provider comparison (AWS/GCP/Azure). 2. GPU vs CPU inference cost analysis. 3. Autoscaling strategy. 4. Model compression strategy. 5. API usage optimization. ------------------------------------------------------- GLOBAL DEPLOYMENT PREPARATION ------------------------------------------------------- 1. Adapt reference ranges per region. 2. Handle unit conversion automatically. 3. Support multilingual medical terminology. 4. Cultural sensitivity in explanations. ------------------------------------------------------- SENSITIVE CASE HANDLING ------------------------------------------------------- 1. Mental health crisis detection. 2. Self-harm risk phrases detection. 3. Abuse or violence indicators. 4. Immediate crisis helpline recommendation when needed. ------------------------------------------------------- SIMULATED PATIENT TESTING ENVIRONMENT ------------------------------------------------------- 1. Create synthetic patient scenarios. 2. Stress-test model on edge cases. 3. Test rare disease detection limits. 4. Evaluate worst-case error impact. ------------------------------------------------------- CODE GENERATION RULES ------------------------------------------------------- When generating code: - Write modular, clean, production-ready code. - Use Python (FastAPI) for backend. - Use React + Tailwind for frontend. - Use PostgreSQL for database. - Use Docker for deployment. - Include environment variable handling. - Include input validation. - Include logging middleware. - Include error handling. - Include unit test examples. Generate code file-by-file. ------------------------------------------------------- AI SELF-IMPROVEMENT LOOP ------------------------------------------------------- Design an iterative loop that: 1. Collects anonymized performance metrics. 2. Evaluates prediction accuracy. 3. Detects drift in model performance. 4. Suggests retraining schedule. 5. Improves calibration over time. ------------------------------------------------------- TRANSPARENCY REQUIREMENTS ------------------------------------------------------- System must always output: - Risk % - Suggested additional tests - Clear medical disclaimer ------------------------------------------------------- ETHICAL CONSTRAINT ------------------------------------------------------- This system is a Clinical Decision Support Tool. It must never be positioned as a replacement for licensed medical professionals. ------------------------------------------------------- FINAL DELIVERABLE STRUCTURE ------------------------------------------------------- Present your work in this order: 1. Executive Summary 2. System Architecture 3. Technology Stack 4. Core Modules Description 5. Data Flow Diagram Explanation 6. ML Strategy 7. Safety Framework 8. API Design 9. Deployment Plan 10. Validation & Testing Plan 11. Regulatory Considerations 12. Future Improvements indian based app
