AWS Builder Center

Building DiagnoFlow: Resilient Clinical Triage Copilot with Amazon Bedrock

#generative-ai, #bedrock, #healthcare, #python, #nextjs

The Problem We Solved

In emergency departments and primary healthcare clinics, doctors and triage nurses make rapid triage decisions under intense cognitive overload. A missed flag during acute coronary syndrome or severe hypoxia can lead to preventable mortality.
We built DiagnoFlow, an intelligent clinical triage copilot that ingests patient vitals and chief complaints to deliver sub-second risk stratification, differential diagnoses with ICD-10 codes, and clinical action directives.

Architecture & How It Works

DiagnoFlow uses a dual-path resilience architecture:
  1. Frontend: Next.js & Tailwind CSS for real-time telemetry inputs and responsive emergency status gauges.
  1. Backend: Python FastAPI managing schema validation and workflow execution.
  1. AI Layer: Integrates Amazon Bedrock via  boto3  ( bedrock-runtime ), invoking Claude foundation models to produce strict, validated JSON medical summaries.
  1. Offline Fallback Engine: If internet or cloud connectivity drops, an offline deterministic scoring engine takes over in under 60ms to eliminate emergency department downtime.

Key Learnings & Engineering Takeaways

  • Designing strict typed Pydantic schemas to safely parse and validate LLM outputs in safety-critical medical contexts.
  • Balancing latency vs. diagnostic depth by testing models on Amazon Bedrock.
  • Building resilient cloud-to-edge fallback workflows for zero-downtime environments.



Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article