BloodLink AI

BloodLink AI leverages LLM chat, ML forecasting, and multi-channel alerts to link patients, hospitals, and donors in real time—ensuring timely transfusions and cutting blood wastage.

Comment

BloodLink AI

Elevator Pitch: BloodLink AI leverages LLM chat, ML forecasting, and multi-channel alerts to link patients, hospitals, and donors in real time—ensuring timely transfusions and cutting blood wastage.

Tagline: BloodLink AI – Your AI-powered lifeline for smarter, faster blood donation.

About the project

Inspiration

The inspiration for BloodLink AI came from witnessing the critical delays and inefficiencies in emergency transfusions during volunteer work at local hospitals. Recognizing how fragmented systems and manual coordination can cost lives, we set out to build a smarter, AI-powered solution.

What it does

BloodLink AI seamlessly connects patients, hospitals, and donors through an intuitive LLM-driven chat interface. It predicts donor availability, orchestrates multi-channel alerts, and dynamically manages inventory to ensure the right blood reaches the right recipient at the right time.

How we built it

  • Frontend: React + TypeScript, Material UI.
  • Backend: Python FastAPI microservices on Azure App Service & AKS.
  • AI/ML: Azure OpenAI (GPT-4) for chat, Azure ML for forecasting, Azure Cognitive Services for document parsing.
  • Infrastructure: Azure AD B2C, Azure SQL, Cosmos DB, Blob Storage; CI/CD via GitHub Actions.

Challenges we ran into

  • Integrating disparate APIs (e-RaktKosh, courier services) with inconsistent data schemas.
  • Optimizing ML model accuracy under limited historical data, avoiding overfitting.
  • Ensuring real-time LLM performance while controlling Azure costs.

Accomplishments that we're proud of

  • Delivered an end-to-end MVP with GPT-4 chat triage in 48 hrs.
  • Achieved forecasting RMSE < 0.15 with cross-validated Azure ML models.
  • Automated surplus redistribution, reducing simulated waste by 30 %.

What we learned

We learned the critical importance of data validation, monitoring, and robust error handling in healthcare applications. For example, our forecasting error metric follows:

$$ \mathrm{RMSE} = \sqrt{\frac{1}{n} \sum_{i=1}^n (y_i - \hat{y}_i)^2} $$

Rigorous testing and continuous retraining maintained accuracy.

What's next for BloodLink AI

  • Integrate real-time courier GPS tracking and ETA predictions.
  • Expand multi-language support using Azure Translator.
  • Implement blockchain-based blood provenance for traceability.
  • Launch pilot programs with partner hospitals and NGOs.

Built With

  • aes-256
  • azure
  • azure-ad-b2c
  • azure-app-service
  • azure-blob-storage
  • azure-cognitive-services-(form-recognizer,translator)
  • azure-communication-services-(sms-&-whatsapp)
  • azure-cosmos-db
  • azure-kubernetes-service-(aks)
  • azure-openai-(gpt-4)
  • azure-sql-database
  • docker
  • facebook-graph
  • fastapi
  • github-actions
  • kubernetes-(aks)
  • material-ui
  • oauth2
  • power-bi-embedded
  • progressive-web-app-(pwa)
  • python
  • react
  • tls
  • twitter
  • typescript
Share this project:

Updates