Last-Drop

🩸Last-Drop

Thalassemia patients need frequent, predictable transfusions; missed matches can be life-threatening.
Donor searches are often last-minute, fragmented across groups, calls, and messages.
We saw an opportunity to use AI to predict donor availability, optimize matching, and coordinate care—while keeping privacy and consent at the center.

Repeat donors matter: patients benefit from compatible, recurring donors who build trust over time.
Demand is uneven across zones and time; festival seasons, exams, and local events affect donor turnout.
Friction kills conversion: even willing donors need simple, timely nudges and clear next steps.
Privacy drives participation: donors and patients are more likely to engage when data usage is transparent and minimal.
Accessibility wins: many users prefer WhatsApp/SMS and voice over app-only flows.

Define the core flows: patient request → AI-recommended donors → confirmation → transfusion status tracking.
Model the signals: donor cooldown, past response behavior, distance/ETA, zone demand, and urgency.
Build a modular system: start with a PWA and messaging (WhatsApp/SMS), add IVR for voice sign-ups, and a coordinator dashboard.
Integrate responsibly: design an adapter for e-RaktKosh/Blood Bridge (mock in hackathon, real post-hackathon).
Bake in privacy: consent ledger, role-based access, encryption, audit trails from day one.
Measure what matters: time-to-match, fill rate, donor retention, and patient satisfaction.

Data access variability → Designed a plug-in integration layer with mock data for the demo; real connectors post-hackathon.
Cold-start in new zones → Began with rules-based matching and coordinator inputs; AI improves as data grows.
Donor fatigue → Added frequency caps, rotation, and best-time-to-nudge logic.
Privacy & consent → Granular consent settings, minimal data collection, clear user controls.
Operational adoption → Simple UI, WhatsApp-first flows, and clear, clinician-reviewed content in multiple languages.

Create a request in the app/PWA (or via coordinator).
See predicted next transfusion date and a prep checklist.
Track progress: “2 donors confirmed, ETA 35 minutes”.
Ask questions via a multilingual care copilot (educational guidance, not medical advice).

Eligibility survey (short, rules-based): Eligible / Defer / Manual Review.
Receive smart, timely nudges (WhatsApp/SMS) with one-tap confirm and center recommendation.
View personal impact stats (e.g., “You supported 6 transfusions this year”) and badges (opt-in).

Live match recommendations ranked by blood group, ETA, reliability score, and zone urgency.
Heatmap of demand/urgency by zone to plan drives and outreach.
Inventory tiles (from e-RaktKosh adapter: mock in demo).
Analytics: conversion funnel, time-to-match, donor retention trends.

Predictive Donor Availability: Learns from past donation patterns, cooldowns, and response behaviors to score likelihood to donate now.
Constraint-Aware Matching: Optimizes matches by blood group, distance/ETA, donor cooldown, zone urgency, and inventory signals.
Zonal Demand & Heatmap: Aggregates requests into geospatial cells to visualize and forecast where and when to recruit.
Nudge Optimization: Tests message timing and channels (WhatsApp/SMS) to improve confirmation rates while respecting user preferences.
Care Copilot (RAG + Multilingual): Provides accessible, clinician-reviewed education for patients and caregivers.

PWA for donors/patients (fast, installable on mobile).
Donor eligibility survey with clear outcomes and consent.
AI-driven matching and zonal heatmap with urgency colors (red/orange/yellow).
WhatsApp/SMS confirm/cancel flows with links and frequency caps.
Voice registration (IVR) in English with DTMF fallback; more languages post-hackathon.
Coordinator dashboard with live queue, match suggestions, and basic analytics.
Mock e-RaktKosh adapter to simulate inventory data and center recommendations.

Faster matches and fewer missed transfusions.
Higher donor retention via respectful, personalized engagement.
Smarter recruitment through zone-level insights and forecasts.
Trust by design through transparent consent, minimal data, and strong security.
Scalable foundations for real integrations with Blood Warriors initiatives.

Live e-RaktKosh and Blood Bridge integrations.
Expanded language support (IVR and in-app).
Learning-to-rank for patient–donor cohorts and extended antigen compatibility when available.
Federated learning pilots to improve models without sharing raw data.
Employer/college donation drives, badges, and certificates to grow the donor base.