Every tier has an independent admin panel, isolated Row-Level Security policies, and configurable modules. A config change at Tier 2 (Group) propagates down β a config at Tier 4 (Department) stays scoped to that department.
TIER 01
Country
Malaysia Β· Canada Β· Singapore
Separate data residency per jurisdiction
Country-level compliance configuration
Local payment rail integration (FPX, Interac, NETS)
Independent admin panel per country
TIER 02
Healthcare Group
KPJ Β· Raffles Β· Columbia Asia
Group-level analytics and consolidated billing
Cross-hospital reporting dashboard
Group admin with hospital management
Isolated RLS policies per group
TIER 03
Hospital
KPJ Damansara Β· KPJ Ampang Β· KPJ Ipoh
Patient check-in flows and kiosk configuration
HIS integration (KCIS, NEHR, MyHealth)
Staff dashboard and queue management
Configurable modules per facility
TIER 04
Department
A&E Β· Cardiology Β· Orthopaedics Β· OB/GYN
Queue management and specialist routing
Acuity threshold configuration per department
Department-level analytics
Independent module activation
CountryHealthcare GroupHospitalDepartment
Each tier is a separate RLS boundary. Zero cross-tenant data access is architecturally enforced β not just policy.
86-Table PostgreSQL Schema
Row-Level Security enforced at every table. Tenant isolation is a database guarantee, not an application assumption. Zero cross-tenant data leakage is possible by design.
The only HIPAA and PDPA-compliant LLM clinical triage solution deployed in ASEAN. No patient data leaves the hospital network.
MODEL
MOVO-X AI Engine
Deployed inside hospital private network with hospital-enforced network policy. API calls routed through hospital's internal gateway β never via public internet path. Zero patient data leaves the hospital.
PROMPT ARCHITECTURE
Structured + Anonymised
System prompt: clinical context (department, acuity thresholds, active language). User turn: anonymised symptom description β no direct patient identifiers sent to model. Prompt construction happens inside hospital network boundary.
OUTPUT
SOAP Note + Acuity Score
Structured output extraction: SOAP clinical note format. Acuity score 1β5 assigned from symptom keywords, duration, and risk factors. Delivered to doctor's dashboard before patient enters the consultation room.
Performance
LLM Response Time< 800ms average
Full Triage Session90 seconds average
Acuity AssignmentReal-time β before queue ticket issued
Multilingual Support7 languages (EN, MS, ZH, TA, FR, ID, TH)
Resilience & Audit
Fallback: if AI unavailable, kiosk switches to structured symptom checklist β no patient blocked
Every triage session logged: prompt, response, SOAP output, acuity score
Timestamp, kiosk ID, and patient encounter ID on every audit record
Session replay available for clinical governance review
Anomaly detection on acuity score distribution per department
Triple-Layer Security
Each layer is independently verifiable. Security is not a policy β it is an architectural constraint enforced at network, application, and database level simultaneously.
LAYER 01
Network Layer
Hospital private network deployment β kiosk never on public internet
Firewall rules with zero inbound ports exposed on kiosk hardware
VPN for remote administration and support access
Network segmentation between kiosk LAN and hospital HIS LAN
All external API calls outbound-only, proxied through hospital gateway
LAYER 02
Application Layer
JWT authentication with 15-minute access token expiry
Refresh token rotation with single-use enforcement
Rate limiting on all API endpoints β per IP, per tenant, per user
OWASP Top 10 compliance β safe DOM APIs enforced across entire codebase
Content Security Policy headers on every response
Input validation with Zod schemas at every API boundary
LAYER 03
Database Layer
Row-Level Security policies enforced on every table β no exceptions
Supabase service role key never exposed to client or browser
All queries use prepared statements via Prisma β no raw SQL from user input
Column-level encryption for PII fields (IC number, NRIC, passport)
Automated daily backups with Point-in-Time Recovery (PITR)
Zero cross-tenant query paths possible by architectural design
18-Month Compliance Head Start
Each jurisdiction's compliance stack was architected from day zero β not retrofitted. Competitors entering ASEAN healthcare cannot replicate this in under two years.
MY
Malaysia
PDPA 2010 β consent management, data minimisation, breach notification
PHFSA 1998 β Private Healthcare Facilities and Services Act compliance
MOH certification pathway and Circular 5/2019 digital health records mandate
Supabase ap-southeast-1 (Singapore) with envelope encryption
KMS in ap-southeast-5 (Jakarta) for MY-adjacent residency
MY-resident Data Protection Officer appointed
CA
Canada
PIPEDA (federal) β Privacy Act digital health alignment
Law 25 (Quebec) β world's strictest provincial privacy law
PHIPA (Ontario) β Personal Health Information Protection Act
OHIP Health Card verification integration
Data residency in Supabase ca-central-1 (Montreal)
Consent audit trail captured per encounter, 7-year retention
SG
Singapore
PDPA (SG) β Personal Data Protection Act compliance
HBRA β Healthcare Bill of Rights Act alignment
NEHR integration β National Electronic Health Record API
HealthHub API for patient-facing data portability
MAS TRM guidelines for financial data and payment handling
MOH Singapore licensing pathway for private hospital operators
Production Infrastructure
Every infrastructure choice is production-verified β not planned. 62 end-to-end tests pass on every pull request. OWASP ZAP security scan runs on every CI build.
Our CTO is available for architecture deep-dives, security reviews, and code walkthroughs. OWASP ZAP scan results, E2E test suite output, and schema documentation are available in the investor data room.