Med-SEAL Suite Documentation
Empowering 1.8 million Singaporeans with chronic diseases to thrive between clinic visits.
The Problem
In Singapore, 1 in 3 adults lives with a chronic condition — diabetes, hypertension, or hyperlipidemia. They see their doctor for 15 minutes every 3–6 months. That leaves 4,300+ hours a year where patients manage medications, diet, symptoms, and anxiety alone.
During those hours, patients:
💊 Forget doses — 50% of chronic disease patients are non-adherent to medication
📉 Miss warning signs — a rising HbA1c or blood pressure trend goes unnoticed until the next visit
😰 Feel unsupported — questions go unanswered, motivation fades, conditions worsen
🏥 End up in A&E — preventable complications from poor self-management cost SGD 2.5B annually
The HealthierSG initiative asks patients to take ownership of their health. But ownership without support is just burden.
The Solution
Med-SEAL (Medical — Safe Empowerment through AI-assisted Living) is an AI-powered health companion that walks beside the patient in those 4,300 hours.
It speaks their language — English, Mandarin, Malay, or Tamil. It knows their medications, conditions, and lab results. It reminds them gently, explains clearly, escalates urgently, and never pretends to be a doctor.
Behind the scenes, 7 specialised AI agents collaborate to deliver proactive, empathetic, culturally-aware chronic disease management — all built on Singapore’s own SEA-LION National LLM and interoperable HL7 FHIR R4 healthcare standards.
“Uncle Lim, 68, Type 2 diabetic. He missed his Metformin twice this week. Med-SEAL noticed, sent a gentle nudge in Mandarin. His blood glucose was trending up — Med-SEAL flagged it to Dr. Tan before Friday’s appointment. At the visit, Dr. Tan had a pre-visit brief ready with 30 days of adherence data, biometric trends, and PRO scores. The consultation was focused, informed, and 10 minutes shorter.”
— This is the experience Med-SEAL enables.
What Med-SEAL Does
For Patients |
For Clinicians |
|---|---|
🗣️ Chat in 4 languages about health concerns |
📋 Auto-generated pre-visit briefs with 30 days of data |
💊 Proactive medication reminders |
🚨 Tiered escalation alerts (low → medium → high) |
📊 Track vitals, labs, and trends over time |
📈 Population-level adherence and outcome dashboards |
🥗 Culturally-aware dietary advice (hawker food, festive meals) |
🩺 CDS Hooks integration in OpenEMR |
📅 Appointment booking, reminders, and preparation |
🔒 Full FHIR R4 audit trail |
⚠️ Emergency detection → immediate 995 guidance |
📱 Caregiver mode with consent-gated access |
How It’s Built
Layer |
Technology |
|---|---|
AI Agents |
7 LangGraph agents orchestrated by a safety-guarded rule-based router |
LLM |
SEA-LION v4-32B (AI Singapore) + Med-SEAL-V1 (fine-tuned clinical model) |
Safety |
Dual-layer guard: 21 regex patterns + SEA-Guard LLM classification |
Data |
HL7 FHIR R4 via Medplum — interoperable with any hospital system |
EMR |
OpenEMR integration with CDS Hooks |
Patient App |
React Native (iOS & Android) |
Infrastructure |
GCP Cloud Run + GKE (Singapore region) |
Overview
Components
AI Platform
Operations & Reference
User Guide