Inspiration Thalassemia sufferers experience frequent, life-threatening transfusion requirements that rely upon rapid, dependable blood logistics. Even with national initiatives such as [e-RaktKosh], in-principle donor mobilization and matching of blood within hospitals remain manual, error-prone, and too slow — particularly in areas with patchy internet or over-stretched clinical teams. We were motivated by the dream of connecting pressing medical requirement and timely resource mobilization so that no child life is lost because of operational inefficiencies. Seeing doctors and NGOs getting frustrated with siloed data, phone calls, and spreadsheets motivated us to create a clinical-grade, AI-powered platform that is for doctors and blood banks, not consumers.
What We Learned
Real impact is when technology is heavily ingrained in hospital processes instead of as a consumer application. Predictive AI, when combined with actual hospital inventories and national registries, can greatly enhance speed and accuracy of blood logistics. Automating time-critical processes (e.g., Thalassemia transfusions and chelation therapy alerts) in hospital systems eliminates mental burdens for stressed clinicians and guarantees high-risk patients are never overlooked. System interoperability (e.g., e-RaktKosh) and data privacy are not amenities — they are requirements for real-world adoption.
How We Built It -Architecture
Multi-purpose web dashboard (React + Tailwind) for physicians, transfusion technologists, administrators, and NGO coordinators. Modular back-end with FastAPI, PostgreSQL for patient and blood stock management, Redis for real-time alerting. -AI Modules Time-series forecasting (using Prophet and LSTM) for forecasting blood stock requirements. Scikit-learn/XGBoost-based donor-patient matcher. Taking into account compatibility, availability, as well as history of donations. Patient transfusion scheduler with automated reminders and risk scoring from clinical laboratory inputs. -Integration & Communication End-to-end integration of e-RaktKosh APIs for donor registration and inventory synchronization. Automated email/SMS notifications to donors and coordinators (using Twilio, SMTP). Cross-hospital low-stock warnings message queue triggers (Kafka/MQTT). -Data Security Role-based access (JWT/OAuth2), audit logging, and AES-256 encrypted patient records. -Deployment On-premise deployment of Docker (for edge hospitals) and optional central web hosting.
Problems We Faced
Messy Data: Cleaning and mapping hospital and blood bank data (often unstructured, missing, or redundant) was a massive task. Interoperability: Compatibility with e-RaktKosh's API standards and data structures required extensive reverse engineering and testing. AI Reliability: Foretelling unusual occurrences (such as sudden spikes in blood need) compelled us to ensemble numerous models for precision. Usability: Creating a system that is powerful yet easy enough for senior hematologists all the way through to hospital volunteers involved many cycles of feedback. Offline-First: Maintaining data integrity and regular sync in low-connectivity environments was accomplished via custom sync logic and distributed DB design.
Built With
- dockerized
- fastapi
- postgresql
- python
- react
- redis/celery
Log in or sign up for Devpost to join the conversation.