• Next.js
• Node.js
• PostgreSQL
• Neo4j
• Redis
• Pinecone
• LangChain
• Gemini
• Ollama
• RAG
• Docker
• AWS ECS

Overview

JeevanSetu is a unified patient-flow network for emergency care. Frontline workers capture vitals and symptoms, run a safety screening check, retrieve clinical guidelines with retrieval-augmented generation, and route patients to nearby hospitals based on live department capacity and specialized infrastructure.

The system is built on human-in-the-loop validation: it acts as a co-pilot for clinicians and never makes a final medical assessment or referral on its own. Every AI recommendation is verified, adjusted, or overridden by a certified doctor before a patient is transferred.

Highlights

• Grounded decisions: RAG over 1,400+ pages of medical documents and guidelines, filtered by department and severity.
• Voice-first intake: hands-free symptom capture so triage nurses can work without leaving the patient.
• Resource-aware routing: Neo4j models hospital departments, specialties, and availability to pick the right facility.
• Human-in-the-loop by design: no autonomous referrals, with an explicit review step in the workflow.

Architecture

Overview

A Next.js 15 (App Router) frontend paired with a Node.js/Express API, deployed as Dockerized services on AWS ECS.

Data & retrieval

• PostgreSQL — primary transactional records (patients, encounters, referrals).
• Neo4j — graph model of hospitals, departments, and specialties for resource-based routing.
• Pinecone — vector store for semantic retrieval over clinical guidelines.
• Redis — caching and event streaming between services.

GenAI pipeline

• Multi-agent LangChain orchestration with distinct agents for intake parsing, safety screening, guideline retrieval, and routing.
• RAG grounding keeps responses tied to retrieved medical documents instead of model memory.
• Hybrid model routing across cloud Gemini and local Ollama models behind one interface.

Human-in-the-loop

• Every recommendation is surfaced as a draft, not a decision; a clinician reviews and confirms before any referral is created.