Synpulse — Agentic AI for Healthcare
Agentic AI for Chronic Disease Management — NUS-Synapxe-IMDA AI Innovation Challenge 2026.
Chronic disease management is a coordination nightmare — patients juggle multiple medications, forget doses, misread symptoms, and fall through the cracks between appointments. Synpulse is an agentic AI system that acts as a 24/7 clinical copilot for patients with chronic conditions like diabetes, hypertension, and asthma.
The system uses a multi-agent architecture: a retrieval agent pulls relevant clinical guidelines and patient history, a reasoning agent evaluates the current situation against those guidelines, a safety agent checks for medication interactions and contraindications, and an output agent drafts a structured recommendation in patient-friendly language.
Built for the NUS-Synapxe-IMDA AI Innovation Challenge 2026, it demonstrates how agentic AI can bridge the gap between clinical knowledge (buried in PDFs and databases) and patient action (simple, timely, personalized nudges). Every recommendation is traceable back to the guidelines that informed it.
- ▸Multi-agent orchestration: Retrieval Agent → Reasoning Agent → Safety Agent → Output Agent. Each agent has a defined role and passes structured data to the next.
- ▸RAG pipeline: clinical guidelines chunked and embedded in ChromaDB, retrieved with hybrid search (semantic + keyword).
- ▸Safety layer runs medication interaction checks against a curated drug database before any recommendation is finalized.