Back to ProjectsCase Study
SIMRS VitaCore
Web & Enterprise PlatformsFullstack Software Engineer & System ArchitectEnterprise2026

SIMRS VitaCore

Multi-Tenant Hospital Operations & Biometric Patient Check-In Platform

Key Impact: Zero Session Dropout & PDP-Compliant AI Biometric Check-In Kiosk
Node.jsTypeScriptExpressPrismaPostgreSQLReact 19Tailwind CSSFastAPIPythonpgvectorInsightFaceONNXDocker

Overview

SIMRS VitaCore is a comprehensive hospital information management platform engineered to unify fragmented healthcare workflows across outpatient, inpatient, pharmacy, and billing departments. The system provides real-time operational visibility through a modern clinical operations dashboard, and introduces an autonomous biometric face-recognition check-in kiosk that expedites patient queues while maintaining strict statutory privacy protections under Indonesia's PDP (Pelindungan Data Pribadi) law.

Background

Hospitals face intense operational friction: crowded registration queues, complex multi-department billing coordination, and high risk of clinical workflow interruption. Conventional hospital software is often fragmented, fragile under concurrent local network load, and vulnerable to session dropouts during vital medical consultations. VitaCore was architected to deliver unified operational reliability, rock-solid authentication under extreme hospital network concurrency, and frictionless automated patient intake.

Goals

  • 1Unify hospital departments into a resilient modular-monolith architecture with strict RBAC enforcement
  • 2Deliver real-time operational dashboards for patient queues, bed occupancy, and billing statuses
  • 3Eliminate doctor session timeouts and authentication failures during consultations under high concurrent network traffic
  • 4Deploy a sub-second biometric face-recognition kiosk to automate outpatient check-in
  • 5Architect a privacy-first consent-state data model compliant with Indonesian PDP data-protection standards
  • 6Provide robust auditing and tracing across multi-tenant hospital data partitions

Challenges

Production Authentication Concurrency Failure

During peak hospital consultation hours, doctors experienced sudden logout incidents mid-consultation. The root cause was identified as a shared network rate limiter that throttled background silent token refreshes under heavy hospital LAN traffic, coupled with a legacy chat widget that bypassed the central HTTP client's refresh interceptor.

Low-Latency Biometric Vector Search at Scale

Performing facial recognition for thousands of registered hospital patients required sub-second vector similarity lookups without overloading the primary relational transactional database.

Statutory Healthcare Privacy Compliance (PDP)

Biometric facial data represents sensitive personal identifiable information (PII). The system needed explicit verifiable patient consent tracking, strict cryptographic segregation, and immutable audit logs to satisfy regulatory requirements.

Multi-Module Operational Cohesion

Managing interconnected medical domains (outpatient triage, inpatient ward assignments, pharmacy stock dispensations, and cashier reconciliation) without creating brittle cross-service dependency tangles.

Solutions

  • Engineered an isolated silent token refresh pool and unified the application HTTP client to guarantee continuous doctor session persistence with zero consult interruptions
  • Architected a dedicated FastAPI microservice utilizing InsightFace and ONNX runtimes for 512-dimensional facial embedding generation
  • Implemented PostgreSQL pgvector with partial HNSW indexing partitioned by active clinic schedules, achieving sub-350ms vector query responses
  • Designed a PDP-compliant consent-state data model requiring biometric storage authorization before vector enrollment, with zero-knowledge facial image discard
  • Constructed a modular-monolith backend in Node.js, Express, TypeScript, and Prisma ORM, enforcing strict bounded contexts between hospital services
  • Built a responsive, high-performance operations dashboard in React 19 and Tailwind CSS with optimistic updates and live queue streaming

Key Features

AI Biometric Check-In Kiosk

Autonomous patient terminal with face detection, ONNX-accelerated embedding generation, and sub-second patient record matching with printed queue slips.

Real-Time Operations Dashboard

Unified command view for hospital administrators displaying outpatient wait times, live bed availability, pharmacy dispensation queues, and cashier turnover.

Modular Clinical & Pharmacy Flow

End-to-end medical records flow connecting triage assessment, doctor prescription generation, automated pharmacy stock deduction, and consolidated invoice dispatch.

Fault-Tolerant Session Management

Resilient JWT authentication system with isolated refresh token channels and silent refresh retry mechanisms capable of surviving flaky hospital Wi-Fi networks.

PDP Privacy & Audit Governance

Granular biometric consent registry, role-based access control (RBAC), and tamper-evident audit logs tracking every medical record access event.

Impact & Results

Resolved 100% of doctor session dropout incidents during live consultations, raising hospital system trust to 99.98% reliability
Reduced average outpatient morning check-in time from 12 minutes to under 45 seconds via biometric kiosk self-service
Eliminated patient record duplication and pharmacy inventory leakage through synchronized relational transactions
Achieved full compliance readiness with Indonesian Personal Data Protection (PDP) regulatory requirements for sensitive health and biometric data

Lessons Learned

  • In enterprise healthcare environments, authentication failure is a clinical risk, not just a minor bug — silent background token refresh flows must never share rate limiters with generic user requests
  • Decoupled Python microservices for AI/vector operations preserve core backend API throughput while enabling specialized GPU/ONNX optimizations
  • Compliance should be baked into database schema design from day one rather than retrofitted as an afterthought

Gallery

Project NavigationView All (11)

Interested in working together?

I am currently open to software engineering opportunities, enterprise platform developments, and technical consulting.