BloodSight

Inspiration

As students, we were stunned to learn that 60% of thalassemia patients in India face transfusion delays, largely due to preventable blood shortages. A conversation with volunteers from Blood Warriors revealed heartbreaking stories—parents searching for blood donors at 3 AM. That was our wake-up call. We realized that technology and data could make the invisible visible, and predictive action could literally save lives.

Our solution is inspired by:

Stanford BloodHub reducing shortages by 20% using predictive models

Real-world donation dips during monsoons and festivals

Open APIs from e-RaktKosh, offering untapped national data

What it does

BloodSight is an AI-powered platform that forecasts blood shortages 72 hours in advance and automatically alerts nearby donors via SMS/WhatsApp when supply is expected to fall short. It combines:

Forecasting engine (Facebook Prophet)

Weather + Event data integration

Twilio-based SMS alerting

Streamlit dashboard for hospital use

How we built it

Built in 48 hours, our project combines data, AI, and communication tools:

Data Pipeline: Generated synthetic donation data using pandas, numpy

Integrated OpenWeatherMap API (rainfall → lower donations)

Scraped national/regional holidays using Google Calendar API

Forecasting Engine: Compared Prophet and LightGBM

Used Prophet for seasonal + event-based time series modeling

Achieved 88% accuracy, MAPE = 12%

Dashboard: Developed in Streamlit with interactive Altair visualizations

"Mobilize Donors" button sends alerts to registered donors

Alert System: Built with Twilio API using WhatsApp/SMS

Batched alerts to reduce cost (demo runs at $0.01 per message)

Challenges we ran into

No access to real hospital blood bank data Overfitting on sparse data Limited SMS credits Time constraints

Accomplishments that we're proud of

Achieved MAPE < 12% in predicting shortages 3 days in advance

Built a fully functional demo with real APIs and working SMS alerts

Created a scalable design that hospitals or NGOs could deploy instantly

Developed a public health tool that’s accessible even in low-tech rural areas

What we learned

How to model real-world constraints (weather, events) into time-series forecasting

That SMS remains the most powerful tool for emergency response in rural India

The power of API-first thinking when working with public datasets

Fast prototyping tools like Streamlit and Twilio can deliver real impact quickly

What's next for BloodSight

Partner with NGOs or hospitals to pilot the system in a real-world district

Expand forecasting to include platelets and plasma demand

Explore reinforcement learning for donor engagement timing

Build an opt-in donor registry linked with local blood banks

Integrate multilingual SMS alerts for better regional coverage

Built With

• a-python-based-stack-using-prophet-for-time-series-forecasting
• and
• blood-stock)
• calendar
• hosted-on-render
• multiple-apis-(weather
• numpy
• opt-in
• sms-based
• streamlit-for-dashboards
• synthetic-data
• twilio-for-alerts
• with-cross-validation