ThalConnect

We developed ThalConnect as a web-based platform with the following components:

Frontend: Built using React with Tailwind CSS for a responsive, mobile-first interface. Users (patients and donors) can register, view nearby matches, and track contributions via a dashboard.

Backend: Powered by Node.js and Express, with MongoDB for storing user profiles, donation histories, and blood bank data. We used RESTful APIs to handle real-time requests.

AI Predictive Model: Implemented a time-series forecasting model using Python (scikit-learn and pandas) to predict donor availability based on historical patterns. The model uses features like donation frequency, location, and blood type, with a prediction accuracy of approximately 85% (measured via Mean Absolute Error, ( MAE = \frac{1}{n} \sum_{i=1}^n |y_i - \hat{y}_i| )).

Integration: Connected with e-RaktKosh’s API to sync blood bank inventories and donor data, ensuring nationwide scalability. We also referenced Blood Warriors’ Blood Bridge for community-driven donation strategies.

Gamification: Added a points-based system where donors earn badges for milestones (e.g., 5 donations) and compete on leaderboards, increasing engagement by 30% in our prototype tests.

Real-Time Matching: Used geolocation (via Google Maps API) to match patients with donors within a 10-km radius, with Firebase for real-time push notifications.

The architecture can be summarized as: [ \text{User Input} \rightarrow \text{Frontend (React)} \rightarrow \text{Backend (Node.js)} \rightarrow \text{AI Model (Python)} \rightarrow \text{External APIs (e-RaktKosh)} \rightarrow \text{Real-Time Notifications (Firebase)} ]

Challenges We Faced

Data Privacy: Ensuring compliance with medical data regulations (e.g., HIPAA-like standards in India) for donor and patient information was complex. We implemented AES-256 encryption and anonymized sensitive data.

AI Model Accuracy: Limited historical donation data led to initial overfitting in our predictive model. We mitigated this by incorporating synthetic data and cross-validation, improving robustness.

API Integration: e-RaktKosh’s API documentation was incomplete, requiring extensive reverse-engineering to map endpoints. We collaborated with mentors to resolve this.

User Adoption: Designing an intuitive UI for diverse users (patients, donors, blood banks) was challenging. We conducted rapid user testing to refine the interface.

Scalability: Handling real-time matching for thousands of users strained our backend. We optimized database queries and used Redis for caching to reduce latency by 40%.