SehatSaathi

Inspiration The primary inspiration behind SehatSaathi was the growing disconnect between patients and healthcare providers, especially in overloaded clinical environments. We realized that patients often struggle to understand complex medical reports, manage their health records, or get immediate preliminary guidance for symptoms. We wanted to build a unified digital bridge using Artificial Intelligence that empowers patients with instant, easy-to-understand health insights, while simultaneously providing doctors with powerful, streamlined tools to manage their practice and communicate with patients.

What it does SehatSaathi is a comprehensive, dual-portal digital healthcare platform.

For Patients, it serves as an intelligent health companion. Patients can use an AI triage system for preliminary symptom checking, upload complex medical reports (PDFs/Images) to receive simplified AI-generated summaries, and generate highly personalized diet plans. It also acts as a secure vault for their medical records and prescriptions.

For Doctors, it functions as a lightweight clinic management system. Doctors can broadcast their live clinic location and availability, chat directly with booked patients, instantly issue digital prescriptions, and upload medical documents directly into a patient's personal record vault.

How we built it We architected the backend using FastAPI (Python) to ensure high-performance, asynchronous API handling, backed by a PostgreSQL database managed via SQLAlchemy ORM.

For the frontend, we deliberately chose to use Vanilla HTML, CSS, and JavaScript. By avoiding heavy frameworks, we ensured the application remains incredibly lightweight and lightning-fast, while still delivering a premium, glassmorphic dark-mode UI with smooth micro-animations.

The core intelligence of the platform is powered by the Google Gemini API, which we integrated to handle complex multi-modal tasks like parsing dense medical PDFs for the Report Analyzer, generating tailored diet plans, and powering the conversational triage assistant.

Challenges we ran into One of the biggest challenges was securely handling the dual-role architecture (Doctors vs. Patients) to ensure strict data privacy and segregation across the same API endpoints.

Additionally, processing and extracting text from varied medical documents (images and PDFs) to feed into the Gemini API required careful prompt engineering to ensure the AI's output was accurate, safe, and easily digestible for a non-medical user, rather than returning generic or hallucinated advice.

Finally, building a state-of-the-art, dynamic user interface entirely from scratch without relying on CSS frameworks like Tailwind or Bootstrap was a massive undertaking in responsive design and state management.

Accomplishments that we're proud of We are incredibly proud of building a truly end-to-end platform. Rather than just building a simple AI wrapper, we built a fully functional ecosystem that tackles triage, diet planning, secure medical record storage, live doctor chatting, and digital prescriptions.

We are especially proud of the Report Analyzer feature, which successfully takes the anxiety out of reading dense lab results by translating medical jargon into simple, actionable insights.

What we learned We vastly deepened our understanding of asynchronous database operations using asyncpg and SQLAlchemy within the FastAPI ecosystem. We also learned advanced prompt engineering techniques to constrain large language models (LLMs) to act as specialized, safe medical assistants rather than unpredictable conversational bots.

What's next for SehatSaathi Computer-Vision Rehab Monitoring: Fully implementing a real-time, browser-based pipeline using MediaPipe to track patient joint angles during physical therapy exercises and provide live form correction. Cross-Feature AI Integration: Automatically injecting the data parsed from uploaded bloodwork and lab reports directly into the AI Diet Planner to generate hyper-personalized nutrition protocols based on actual biometric data. Mobile Expansion: Wrapping the platform into a native mobile application to leverage device-level features like push notifications for the Medicine Adherence Tracker.

Built With

• css3
• fastapi
• git
• google-gemini-api
• html5
• javascript
• jwt
• postgresql
• python
• render
• sqlalchemy
• sqlite
• uvicorn