MOVO-X Singapore Technology Architecture
A deep-dive into the engineering behind MOVO-X Singapore: on-premise AI triage with zero PHI egress, polyclinic-scale queue prediction, FHIR R4 NEHR integration, SingPass MyInfo check-in flow, and a three-layer security architecture built for PDPA SG and HCSA 2020 compliance.
AI Triage Architecture
On-Premise · Zero PHI Egress · SET AcuityMOVO-X Singapore's AI triage architecture ensures full PDPA SG compliance: no patient health information leaves the hospital or polyclinic's Singapore infrastructure.
Edge Layer (On-Premise): A quantized clinical LLM processes patient-reported symptoms at the kiosk. The model runs entirely on local hospital hardware — no network connectivity is required for the triage decision itself. Output: a Singapore Emergency Triage (SET) acuity score (P1–P5) and a structured symptom cluster. SET P1 and P2 cases trigger immediate nursing notification via WebSocket push to the staff dashboard.
Cloud Layer (PHI-Free): Symptom category (not raw text), acuity score, and consultation type are sent to Anthropic Claude API to generate enriched clinical language for the doctor's pre-consultation note. Because patient identifiers are stripped before this API call, the cloud processing satisfies PDPA SG's purpose limitation requirement. The enriched note is returned to the clinician dashboard within 3 seconds.
Multilingual Triage Input: Singapore patients can enter symptoms in English, Mandarin, Malay, or Tamil. The on-premise model includes a translation layer that normalises all input to English clinical terminology before symptom classification. This ensures consistent acuity scoring regardless of input language.
Audit Trail: Every triage event is immutably logged with timestamp, model version, acuity score, input language, and clinician override (if any). This log satisfies MOH's clinical audit requirements and PDPA accountability principle. The consent disclosure for AI-assisted triage is displayed in the patient's preferred language before symptom entry.
Queue Management Algorithm
AI · Polyclinic-Scale · 30s UpdatesMOVO-X Singapore's queue engine is purpose-tuned for Singapore's polyclinic and public hospital scale — facilities handling 500–2,000 patient interactions per day.
Queue Assignment Logic: Patients receive a queue position using a weighted scoring function: appointment type (booked GP panel vs. walk-in) × CHAS subsidy class routing × SET acuity weight × arrival time. Walk-in to booked ratio is configurable per time slot. Subsidized patients (Class C/B2) and private patients (Class B1/A) are routed to separate queues with independent service counters.
Prediction Model: A clinic-specific XGBoost regression model predicts wait times using: (1) rolling 30-day consultation duration distribution for each doctor, (2) current queue depth per service type, (3) time-of-day seasonality index for this specific polyclinic, (4) CHAS verification time variance (±90 seconds per patient). Model is retrained weekly on each polyclinic's own data — predictions are clinic-specific, not generic industry averages.
Supabase Realtime Push: Queue state is maintained in Supabase PostgreSQL. Every patient event (check-in, call, no-show, consultation end) triggers a Supabase Realtime broadcast to: (1) TV queue displays — refreshed in under 200ms; (2) Patient's registered mobile number (WhatsApp or SMS) when they are 3 positions from being called; (3) Clinician dashboard showing next 5 patients with AI triage summaries pre-loaded.
Polyclinic Load Balancing: For polyclinic groups (SingHealth, NHGP), MOVO-X provides cross-polyclinic wait time visibility, allowing operations managers to see which nearby polyclinics have shorter queues and optionally suggest patients redirect. This inter-facility intelligence layer is a Singapore-specific capability.
NEHR FHIR R4 Integration and SingPass MyInfo
NEHR · FHIR R4 · SingPass · HCSA 2020MOVO-X Singapore fulfils two critical Singapore-specific integration requirements: NEHR contribution (mandatory under HCSA 2020) and SingPass MyInfo patient identity (96% adult adoption).
SingPass MyInfo Check-In Flow: 1. Patient opens SingPass app and scans MOVO-X kiosk QR code. 2. Patient reviews and approves attribute sharing (name, NRIC, DOB, contact details). 3. SingPass issues a consent token; MOVO-X exchanges it for MyInfo attributes via GovTech API. 4. MOVO-X matches NRIC against appointment records in the cluster HIS. 5. Patient is presented with their appointment details, payment summary, and PDPA consent statement. 6. Total check-in time: under 60 seconds.
The SingPass integration runs over HTTPS with GovTech's production environment. All tokens are single-use and expire after 5 minutes. MOVO-X does not store the SingPass access token — only the resolved MyInfo attributes (which are personal data covered by PDPA consent captured at step 5).
NEHR FHIR R4 Submission: After each consultation, MOVO-X submits two FHIR resources to NEHR via the IHiS API gateway: - Encounter resource: Clinical encounter details (admission type, department, attending provider NPI, duration, discharge disposition). - DocumentReference resource: AI-generated SOAP note linked to the Encounter, formatted per NEHR DocumentReference profile.
Submission uses SMART on FHIR OAuth2 with per-facility client credentials issued by IHiS. Failed submissions (network errors, NEHR downtime) are queued with exponential backoff and retried for up to 72 hours. Submission status is surfaced on the CLINIC_ADMIN dashboard with per-patient NEHR acknowledgement tracking.
Cluster HIS Integration: MOVO-X maintains three integration paths: - SingHealth → TrakCare FHIR R4 API (Patient $match, Appointment read, Encounter write) - NUHS → i.MedOne FHIR R4 + HL7 v2.5 ADT translation layer - NHG → iHIS Sunrise XA HL7 v2.5 with FHIR R4 NEHR passthrough
Security Architecture: JWT, RLS, and PDPA SG Compliance
JWT HS256 · RLS · PDPA SG · ap-southeast-1MOVO-X Singapore's security architecture provides three independent isolation layers, ensuring tenant separation, PHI protection, and full PDPA SG and HCSA 2020 compliance.
Layer 1 — Application (JWT): Every API request carries a clinicx_token cookie: HS256-signed JWT containing clinicId (facility-level), userId, role, and expiry. Verified in Next.js middleware via jose before any handler executes. Kiosk devices authenticate with a separate KIOSK_API_KEY + HMAC signature, scoped to their registered clinicId. No kiosk can impersonate another facility.
Layer 2 — Database (RLS): Supabase PostgreSQL has Row-Level Security on every clinic-scoped table. Policies deny ALL to anon and authenticated Supabase roles. Only the server-side service_role key bypasses RLS — this key is never exposed to client code. Even if the Supabase anon key were leaked, no patient data is accessible.
Layer 3 — Network (Singapore residency): Vercel Fluid Compute in sin1 (Singapore) processes all compute. Supabase ap-southeast-1 (Singapore AWS region) hosts all data. TLS 1.3 enforced for all connections; HSTS max-age=63072000 prevents protocol downgrade. No patient data transits outside Singapore.
PDPA Consent Gate: The kiosk displays a consent notice in the patient's selected language (EN/ZH/MS/TA) before collecting any personal data. The consent notice discloses: data controller identity (facility + MOVO-X), purpose of collection, data sharing (NEHR submission), retention period, and patient rights. Consent is logged as an immutable event with PDPA version reference. This satisfies PDPA Sections 13–15 (Consent Obligation).
DNC Registry Check: Before sending appointment reminders to unverified phone numbers, MOVO-X performs a DNC registry check via the PDPC API. Verified SingPass numbers (MyInfo-sourced) are exempt from DNC as they represent active consent. Marketing communications never go to DNC-registered numbers.
Breach Response: MOVO-X Singapore's breach response procedure targets PDPC notification within 24 hours of internal detection (ahead of the mandatory 3-calendar-day deadline). The CLINIC_ADMIN receives automated alerts via Resend email when anomalous data access patterns are detected by the PostgreSQL audit trigger.
Architecture Summary
Frequently Asked Questions
How does MOVO-X Singapore AI triage work without sending patient data outside Singapore?
MOVO-X deploys an on-premise AI triage model within the hospital or polyclinic's local network. Patient symptom data is processed entirely within Singapore's network — no PHI transits to external APIs. The on-premise model produces a SET (Singapore Emergency Triage) acuity score and a structured symptom summary. Post-triage, clinical language enrichment uses Anthropic Claude API with PHI stripped (symptom category and acuity score only, not patient identifiers), satisfying PDPA SG purpose limitation requirements.
How does MOVO-X integrate with NEHR Singapore?
MOVO-X Singapore integrates with the IHiS NEHR API to submit FHIR R4 Encounter and DocumentReference resources after each clinical visit. The NEHR submission is triggered automatically when the clinician marks the encounter as complete, pulling structured data from the MOVO-X clinical record. Authentication uses SMART on FHIR OAuth2 with IHiS-issued credentials per facility. All NEHR submissions are logged with submission ID, timestamp, and acknowledgement status for HCSA 2020 audit compliance.
How does the SingPass MyInfo kiosk check-in flow work?
The SingPass MyInfo check-in flow: (1) Patient opens SingPass app on their smartphone; (2) Patient taps "Scan QR" and scans the MOVO-X kiosk QR code; (3) Patient selects which MyInfo attributes to share (name, NRIC, DOB, contact); (4) SingPass returns a consent token to MOVO-X; (5) MOVO-X exchanges the token for patient attributes via GovTech MyInfo API; (6) Patient record is auto-populated and appointment confirmed. Total check-in time: under 60 seconds. Fallback: manual NRIC entry for patients without SingPass.
How does MOVO-X handle Singapore's three hospital clusters?
MOVO-X Singapore maintains cluster-specific HIS integrations: TrakCare FHIR R4 for SingHealth (SGH, KKH, CGH, SKH), i.MedOne HL7/FHIR for NUHS (NUH, AH, NTFGH), and iHIS/Sunrise XA for NHG (TTSH, KTPH). Each cluster integration is configured independently but surfaces unified data to the MOVO-X analytics dashboard. Cross-cluster reporting is available for enterprise customers (e.g., healthcare groups operating across clusters).
What is the MOVO-X queue management algorithm for Singapore polyclinics?
MOVO-X's polyclinic queue algorithm accounts for Singapore-specific patterns: MyInfo-verified walk-in ratios, subsidized vs private class routing, CHAS subsidy verification speed, multi-language consultation time differentials, and GP panel appointment integration. The gradient-boosted prediction model is retrained weekly on each polyclinic's own historical data for clinic-specific accuracy. Prediction updates fire every 30 seconds via Supabase Realtime to TV displays and patient mobile apps.
Is MOVO-X Singapore PDPA SG and HCSA 2020 compliant?
Yes. MOVO-X Singapore was designed PDPA-first. All patient data is stored in Supabase ap-southeast-1 (Singapore) — never leaving Singapore jurisdiction. The kiosk consent gate captures PDPA-compliant consent before data collection, with purpose disclosure in all four official languages. HCSA 2020 compliance is achieved through mandatory NEHR submission after every encounter. A dedicated DPO is appointed, and breach response procedures follow the 3-calendar-day PDPC notification requirement.
Deploy MOVO-X in Your Singapore Hospital
PDPA SG · HCSA 2020 · NEHR FHIR R4 · SingPass MyInfo · PayNow + NETS · sin1 region
Request a Demo