ThalAssist- AI-Driven Blood Donor Matchmaking & Predictive

Inspiration The journey began when I learned more about the daily battles faced by Thalassemia patients in India—especially the immense challenge of securing reliable, recurring blood donations. Seeing how the current donor management systems are fragmented and mostly reactive, I was motivated to build a platform that leverages AI to make these connections smarter, faster, and more scalable. The blend of social impact, technology, and real-world healthcare needs inspired me to pursue this idea.

What I Learned Real-World Data Complexity: Working with healthcare and donor data required careful data modeling, awareness of privacy regulations, and robust validation procedures.

ML for Social Good: Implementing time-series forecasting for donor availability showed me the potential—and limitations—of ML in dynamic, real-life scenarios.

API-First Thinking: Building API integrations (with mock interfaces for platforms like e-RaktKosh) highlighted the value of modular, interoperable platforms in the public health domain.

User-Centered Design: Extensive feedback from potential users (patients, donors, NGOs) underscored the need for clear UI/UX, regional language support, and a privacy-first approach.

How I Built the Project Architecture: Designed a full-stack, API-oriented web platform using a microservices architecture for scalability and rapid development.

Machine Learning: Built donor availability models using Python and scikit-learn to forecast optimal notification times for past donors based on their history.

Frontend: Developed the user-facing app with Next.js and React for a progressive, mobile-first experience.

Backend: Used FastAPI with Python for RESTful APIs, handling AI logic, authentication, user management, and integration logic.

Cloud & DevOps: Deployed services on Azure for compute and storage; used Docker for containerization and GitHub Actions for CI/CD.

Database: Utilized MongoDB for flexible data storage (users, donors, campaigns) and Redis for handling real-time, ephemeral data such as donor radar statistics.

Security: Implemented encrypted communications, user consent flows, and detailed audit logging.

Challenges Faced Data Privacy and Consent: Ensuring compliance with privacy laws while enabling impactful AI predictions was a technical and ethical challenge.

Integration Barriers: Simulating integration with platforms like e-RaktKosh and hospital databases required creativity and robust API mocking.

Resource Constraints: Building AI features and a polished front end with demo-only/mock data required focused planning and prioritization.

User Trust: Encouraging users to share sensitive health data meant designing transparent, explainable AI modules and robust privacy safeguards.