Dispatch

Make prescriptions easy and accessible for everyone at any time, and everywhere. Save time for healthcare specialist in a system where your case saves lives.

Comment

Dispatch

Inspiration

Our inspiration comes from applying our concepts from web infrastructure to the healthcare system in a responsible and ethical manner. Our system allows handling a high volume of patients with limited resources and time.

What it does

  • Batches similar patient intakes into cohorts so one clinician action can resolve many cases.
  • Presents a live 2D decision graph for providers that visualizes forks, patient counts, and node state (open/resolved).
  • Provides a patient mobile-focused intake flow, question answering, and status/history.
  • Enforces province/tenant separation and audit logging for regulated actions.

How we built it

Stack and architecture

  • Backend: FastAPI (Python) with MongoDB; entrypoint: backend/main.py
  • Frontend: Expo (React Native + TypeScript) mobile-first UI in dispatch/ (file-based routing)

Challenges we ran into

  • Designing safe cohorting rules that never batch unsafe or province-mismatched patients.
  • Visualizing a large decision tree in a clear, interactive way on web/mobile.

Accomplishments we're proud of

  • Built a full-stack prototype (backend + Expo app) in the existing repo that demonstrates cohorting, provider approval/signature, and patient-facing intake/history.
  • Implemented a robust bulk seeder that generates realistic demo cohorts to stress-test the provider graph and queue UX.
  • Added safety-first guardrails: emergency detection, province isolation, and explicit provider signature for regulated outputs.

What we learned

  • Generative AI accelerates prototype iteration but needs strict schema validation and human review to be safe in healthcare flows.
  • Good UX for providers requires carefully surfacing why an action is recommended (impact, confidence, tag match).

What's next

  • Finish EasyPrescription PDF rendering and individualized batch-sign PDFs.
  • Add more analytics dashboards for time-saved claims.
  • Harden multi-provider claiming, optimistic locking, and audit trail for production-readiness.
  • Expand AI safety classification and provider-style learning while preserving human-in-the-loop requirements.

Known limitations

  • Some provider workflows (batch PDF generation, signature persistence) are not implemented for implementation.
Share this project:

Updates