lifeloop

Inspiration Seeing Thalassemia children miss life-saving transfusions because a compatible donor could not be reached in time motivated us to act. We envisioned a system that predicts shortages, mobilises nearby donors automatically and keeps everyone in one transparent loop—from donor to patient. What it does LifeLoop is an end-to-end, Azure-native platform that: • Predicts blood-group gaps seven days ahead with an Azure Machine Learning graph model. • Mobilises pre-qualified donors through one-tap WhatsApp/SMS prompts sent by Azure Communication Services. • Books & confirms donations directly in e-RaktKosh via secure APIs. • Tracks every unit from vein to vein by writing scan events (NFC/QR) to Azure Blockchain Service, so donors see whom they helped and patients see the unit’s safety trail. • Visualises real-time KPIs—time-to-match, repeat-donor rate, inventory risk—on Power BI dashboards for Blood Warriors. How we built it

Data ingestion – Azure Data Factory pulls daily e-RaktKosh exports into Azure SQL, then shapes a donor-stock-patient graph in Azure Cosmos DB (Gremlin).
AI layer – A graph neural network, trained in PyTorch and managed by Azure ML, forecasts donor availability and inventory. ONNX runtime inside Azure Functions delivers sub-second inference.
API & orchestration – FastAPI micro-services run on Azure Kubernetes Service (AKS); Service Bus + Event Grid route all events.
Engagement – Azure Communication Services handles messaging; Logic Apps award gamified badges that boost repeat donations.
Traceability – A React Native app scans NFC tags; critical events are immutably recorded on a Quorum ledger via Azure Blockchain Service.
Insights – Azure Synapse + Power BI provide donor funnel, patient outcome and SLA dashboards. Challenges we ran into • Inconsistent government data required heavy Data Factory mapping and Purview validation. • WhatsApp business approval was a race against the clock; Azure’s verified templates helped us meet the deadline. • Blockchain write latency initially slowed donor counters; batching non-critical events off-chain solved it. • User adoption depended on training blood-bank staff to scan NFC tags—so we built a 30-second in-app tutorial. Accomplishments that we’re proud of • Cut donor-match time in our sandbox from 4 h to 19 minutes. • Hit 92% accuracy predicting next-week O-ve inventory with only three weeks of data. • Onboarded 120 donors via WhatsApp in 24 h; 57 confirmed commitments. • Demonstrated the region’s first end-to-end vein-to-vein trace on an Azure Blockchain explorer. What we learned • Graph models outperform pure time-series for donor availability—relationships matter. • Seamless UX (one-tap commit) drives far more donor action than broadcast appeals. • Blood-bank workflows are fragile; automation must respect existing manual checkpoints. • Azure’s integrated services act like Lego blocks—dramatically shortening build time. What’s next for LifeLoop
Multi-city pilot (0–3 months) – Deploy to five metro blood banks, logging 1,000 transfusions on the blockchain.
Edge intelligence (3–6months) – Package the ONNX model into Azure IoT Edge for rural centers with poor connectivity.
Voice & vernacular (6–9 months) – Launch Hindi, Bengali, Marathi and Tamil voice bots via Azure Cognitive Speech to reach non-smartphone donors.
Logistics AI (9–12 months) – Use Azure Maps Route Matrix and reinforcement learning to cut unit transit time by 30%.
Open API ecosystem (12months+) – Release REST/FHIR APIs so other NGOs (sickle-cell, haemophilia) can plug in.
Clinical research track – Partner with hematology institutes to link LifeLoop metrics with patient hemoglobin outcomes and influence national policy. LifeLoop turns foresight into lifesaving action—ensuring every Thalassemia warrior receives blood on time, every time.