Arogya-AI: When Ancient Ayurvedic Wisdom Meets Modern Artificial Intelligence

The Arogya-AI project aims to resolve the contradiction between the 'black box' problem of modern AI medical models and the difficulty of scaling traditional Ayurvedic medicine. By combining the precise predictive ability of machine learning with the reasoning ability of large language models, and deeply integrating Ayurvedic constitution theory, it builds a dual-engine intelligent medical diagnosis system. This system provides interpretable and personalized health management solutions for doctors and patients, positioning itself as a clinical decision support system to assist medical professionals.

At a time when AI is sweeping the medical field, modern AI models often lack interpretability due to their 'black box' characteristics. While traditional medicines like Ayurveda emphasize a holistic view and personalization, they face challenges in large-scale application. Arogya-AI was born precisely to address this core contradiction, exploring the integration path between traditional wisdom and modern technology.

Dual-Engine Architecture

1. Deterministic Machine Learning Engine: Based on the Random Forest algorithm, it uses TF-IDF to process symptom text, outputs disease probability and confidence, ensuring mathematical rigor and repeatability.
2. Generative Large Language Model Engine: Adopts Google Gemini 2.5 Pro, combines multi-dimensional information such as patient constitution (Vata/Pitta/Kapha), and generates holistic Ayurvedic regimen plans.

Interpretability and Safety Design

• Explainable AI (XAI) technology generates an 'AI X-ray' to display key decision-making symptoms;
• When confidence is below 35%, mark it as 'insufficient data', and filter harsh disease names on the patient side to avoid panic.

Data Privacy and Permissions

• Clinic ID isolation mechanism simulates HIPAA compliance;
• Role-Based Access Control (RBAC): Doctors have access to the diagnosis dashboard, while patients can view health logs.

Tech Stack

Frontend: React.js + Vite + Tailwind CSS + Framer Motion; Backend: FastAPI deployed as Vercel serverless functions; Data layer: Google Firestore; Authentication: Firebase Auth; Model training: Scikit-Learn + SMOTE to solve class imbalance issues.

• Model Accuracy: Random Forest 100%, Logistic Regression 99.64%, SVM 94.17%;
• Disease Coverage: Supports diagnosis and Ayurvedic treatment plans for 399 diseases, covering multiple categories such as infectious diseases, metabolic diseases, and respiratory diseases;
• Treatment Plans: Includes 50+ traditional herbs (Sanskrit + English names), 30+ treatment procedures, and personalized diet guides;
• Offline Mode: When the network is unavailable, it degrades to the local database, maintains 100% ML prediction accuracy, and provides offline suggestions.

Arogya-AI is clearly positioned as a clinical decision support system, intended to assist rather than replace licensed medical professionals. By fusing ancient wisdom with modern technology, it demonstrates a humanized path for AI in the medical field—an interpretable intelligent assistant that respects individual differences, providing an example for research on medical AI ethics, interpretability, and the modernization of traditional medicine.

The project roadmap includes:

1. Introduce a continuous learning mechanism to optimize the model through doctor feedback;
2. Integrate wearable devices (Apple Watch/Fitbit) to obtain real-time vital signs;
3. Add multilingual support to serve rural patients;
4. Develop a telemedicine and appointment system, supporting video call triggers.