🩸 Blood Bridge AI – Intelligent Donor-Patient Matching System
💡 Inspiration
In India, thousands of thalassemia patients rely on regular blood transfusions to survive. But finding suitable donors—especially in emergencies—is still a manual and stressful process. We spoke to volunteers and learned that WhatsApp groups and Excel sheets are still the main tools used by NGOs like Blood Warriors to locate donors.
We wanted to build a system that uses AI to bridge the gap between patients and donors—fast, smart, and fair.
⚙️ What It Does
Blood Bridge AI is an AI-powered platform that:
- Predicts which blood donors are most likely to be available
- Matches them to urgent patient requests based on compatibility, location, and engagement history
- Sends smart alerts via SMS or WhatsApp using Twilio API
- Helps NGOs visualize top donor matches with a simple interface
🛠️ How We Built It
We built the solution as a modular pipeline:
🧠 ML Model: Trained a logistic regression model (also tested XGBoost) to predict donor availability using:
- Time since last donation
- Distance from patient
- Blood type compatibility
- Donor’s past response rate
💬 Matching Engine: Ranks top donors based on predicted availability and patient location.
🌐 Backend: Built with FastAPI, serving prediction results and donor data.
🖥️ Frontend: A Streamlit dashboard to input patient info and view top matching donors on an interactive map.
📲 Communication: Integrated Twilio API to auto-send messages to selected donors.
☁️ Cloud: Deployed on Azure Web App, using Azure Blob Storage for mock donor profiles.
🚧 Challenges We Ran Into
- ⚠️ No access to real donor data due to privacy—so we had to create realistic synthetic data and simulate behavior patterns.
- 📍 Location data was rough—many donors don’t share GPS, so we used PIN code clustering.
- 🧪 Fine-tuning ML models for both performance and fairness was tricky, especially avoiding urban bias.
- ⏱️ Building a working backend, frontend, and ML pipeline in a short time frame was a big push!
🏆 Accomplishments That We're Proud Of
- Built a fully working ML-based donor prediction system that’s fast and realistic.
- Achieved 87% test accuracy and AUC of 0.91 using synthetic data.
- The Twilio-based messaging system works seamlessly, enabling direct communication with matched donors.
- We created a tool with real social impact potential—not just a prototype.
📚 What We Learned
- How to build AI for real-world impact, especially in healthcare.
- How to think beyond accuracy—context, ethics, and usability matter more.
- Improved skills in FastAPI, Twilio, and Streamlit under time constraints.
- We now understand the real bottlenecks NGOs face, which go beyond just “tech problems.”
🔮 What’s Next for Blood Bridge AI
- 🔐 Add donor consent and verification layers (like Aadhaar-based opt-in).
- 📈 Integrate with e-RaktKosh, hospital databases, and real NGO platforms.
- 🗣️ Add voice-based chatbot or IVR system for rural patients who don’t use smartphones.
- 🧠 Explore advanced ML models to personalize donor ranking even more.
- 🌍 Partner with NGOs to deploy this solution in the real world.
