About Expertise Frameworks Experience Case Studies Services Skills Contact →

Architecture that moved KPIs.

Each case follows the same architectural narrative: the challenge, my role, the frameworks applied, the architecture, the trade-offs I made and why, how it was governed, and the business outcomes — clinical, operational, and financial.

The numbers, up front.

12%
ICU LOS Reduction
Predictive analytics & care-gap alerting integrated into the EDW supported proactive outreach.
85%
HIMSS Stage 6 Readiness
End-to-end EHR/EMR, RIS/PACS, LIS, perioperative, ICU, and portal flows aligned with HIMSS standards.
25%
Faster Month-End Close
UML-modeled revenue-cycle workflows streamlined order lifecycle and charge capture.
40%
Less Admin Overhead
Patient-centric Java/Spring Boot application automated appointment & insurance workflows.
50%
Throughput Boost
Kubernetes-based microservices scaling lifted e-commerce transaction throughput at peak load.
30%
Faster HR Cycles
Cloud-native automation reduced HR leave-approval time via Docker + Java workflows.
90%
ICT Infra Delivered
Baha Medical Center critical infrastructure including network protocols and secure data systems.
$500K+
Cost-Savings Identified
Automated market analysis dashboard delivered real-time insights for stakeholder decisions.

A closer look.

Four engagements where strategy, architecture, and engineering came together. Financial figures marked “modeled” are illustrative estimates — replace with verified client numbers before publishing.

Case 01 · Healthcare · Greenfield Enterprise Architecture

GAD Heart & Lung Institute — Greenfield Digital Health EA

85%HIMSS Stage 6 readiness
7+3departments + clinics
<1%RTO failover
100%BDAT layers modeled
// Logical view · BDAT (ArchiMate-aligned)BUSINESSPatient JourneyRevenue CycleClinical OpsAPPLICATIONEHR/EMRRIS/PACSLISICUPortalDATAFHIRHL7 v2SNOMED/LOINCODS → EDWTECHNOLOGYMirth BusKubernetesVLAN/DRIAM
The challenge

The first specialized cardiac & pulmonary hospital in the Middle East was being built from the ground up. There was no legacy estate to migrate — but also no reference to anchor on. Every clinical, application, data, and technology decision had to be made deliberately and aligned to a single target architecture.

My role

I led the enterprise architecture end-to-end as Healthcare IT Project Engineer and active PMO member — owning the BDAT model, the interoperability design, and the technical review of vendor proposals against the hospital's vision.

Approach & frameworks
  • TOGAF ADM Phases A→D to set vision, business, data, application and technology architectures.
  • ArchiMate to model the patient journey across all layers as one navigable map.
  • HIMSS Stage 6 as the maturity yardstick for every clinical flow.
  • NORA & HIPAA/GDPR as the governance and compliance backbone.
Key trade-offs & decision logic — why this, not that
DecisionConsideredSelectedWhy
Integration Point-to-point interfaces Central integration bus (Mirth Connect)chosen An N×N web of point-to-point HL7 feeds is cheap to start but its cost and fragility grow quadratically. A central bus added upfront effort but cut per-interface cost, isolated failures, and made every new system a single connection — the right TCO over a multi-year estate.
Interoperability HL7 v2 only FHIR-first, HL7 v2 where requiredchosen FHIR future-proofs APIs and portal/IoT integration; HL7 v2 was retained only where vendor systems mandated it. This balanced modern interoperability against vendor reality.
EHR Build custom EHR Buy & integrate certified EHRchosen A greenfield tempts a custom build, but a certified EHR de-risks compliance and time-to-go-live. Architecture effort went into integration and data, not reinventing the record.
Governance & risk
TOGAF ADM gatesArchitecture Review BoardRisk registerHIPAAGDPRZero Trust segmentationAAA · audit loggingEncryption-at-rest
Business outcomes & ROI / TCO

85% HIMSS Stage 6 readiness with main flows mapped per system and department. The bus-based design is projected to reduce integration OPEX versus point-to-point as systems are added — a modeled 30–40% lower cost-per-interface over the estate lifecycle. DR architecture delivered sub-1% RTO failover, protecting clinical continuity.

User & citizen centricity

Patients gained a single portal for access and results; clinicians worked from one coherent record instead of swivel-chairing between systems — shortening the path from arrival to care.

Case 02 · Infrastructure · ICT Healthcare Environment

Baha Medical Center — Secure ICT Deployment

90%critical infra delivered
75%on-schedule completion
0unsegmented zones
100%protocol coverage
// Network & resilience viewCLINICAL SERVICESHIS / ERPImagingLabsAPPLICATIONActive DirectoryDNSEmailWi-FiNETWORKAruba SwitchingPalo Alto FWVLANs · OSPFRESILIENCEHA LinksDR SiteBackups
The challenge

A hospital cannot open without a clinical-grade network. Baha Medical Center needed its entire ICT backbone — switching, firewalls, identity, DNS, and resilience — designed, built, and commissioned to support HIS/ERP rollout on a fixed schedule.

My role

As IT Project Engineer I designed and supervised the implementation of the ICT environment and rolled out the network-protocol layer across the clinical backbone, providing on-site and remote support through commissioning.

Approach & frameworks
  • Segmented, defense-in-depth network design (VLANs, OSPF, IP access lists).
  • Aruba switching and Palo Alto firewalls as the standardized stack.
  • Active Directory + DNS for identity and name resolution.
  • Test-and-commission discipline tied to the project schedule.
Key trade-offs & decision logic — why this, not that
DecisionConsideredSelectedWhy
Resilience Active-active clustering Active-passive clusteringchosen Active-active maximizes utilization but doubles operational complexity and split-brain risk. For a new clinical site, active-passive gave predictable failover and far simpler operations — resilience the team could actually run.
DR Cloud-based DR On-prem DR sitechosen Regulatory and latency constraints for clinical data favored an on-prem DR posture at this stage, with a documented path to hybrid later.
Vendors Best-of-breed mix Standardized Aruba + Palo Alto stackchosen A single, well-supported stack lowered maintenance OPEX and training burden versus a best-of-breed patchwork — the right call for a lean hospital IT team.
Governance & risk
Change controlNetwork segmentationZero Trust principlesAccess listsCommissioning sign-offVendor SOW review
Business outcomes & ROI / TCO

90% of critical infrastructure delivered with full network-protocol coverage and secure patient-data systems, hitting 75% completion on schedule. The standardized vendor stack is projected to lower ongoing maintenance OPEX and shorten mean-time-to-repair — a modeled reduction in support effort for the in-house team.

User & citizen centricity

Clinical staff received a reliable, segmented network from day one — uptime and security that users never had to think about, which is exactly the point.

Case 03 · Application · Patient-Centric Platform

Patient-Centric Healthcare Application

40%less admin overhead
95%production-ready
API-first design
24/7self-service
// Solution view · microservicesBUSINESSAppointmentsInsuranceFront DeskAPPLICATIONSchedulingInsurance SvcNotificationsNode UIDATAPostgreSQLEvent LogTECHNOLOGYSpring BootDockerAPI Gateway
The challenge

Front-desk staff were drowning in manual appointment and insurance workflows — slow, error-prone, and a poor patient experience. The hospital needed an automated, patient-facing platform that operations could trust.

My role

I designed the solution architecture — domain decomposition, API contracts, data model, and the OODA-driven analysis from requirements to use cases to notation — and guided the build to production-readiness.

Approach & frameworks
  • Domain-driven microservices in Java 17 / Spring Boot.
  • API-first contracts with a Node.js front-of-house layer.
  • OODA method: requirements → use cases → notation → build.
  • PostgreSQL data modeling with auditability built in.
Key trade-offs & decision logic — why this, not that
DecisionConsideredSelectedWhy
Decomposition Monolith Microserviceschosen A monolith ships faster on day one, but the roadmap demanded independent scaling of scheduling vs. insurance and parallel team delivery. Bounded microservices traded a little upfront overhead for scalability and change-isolation — justified by the growth plan, not by fashion.
Stack Node.js end-to-end Java/Spring core + Node UIchosen Java/Spring brought transactional rigor and a mature ecosystem for the core domain; Node served the responsive front-of-house. Right tool per layer — a technology-agnostic call.
Scheduling SaaS scheduling product Build on owned serviceschosen An off-the-shelf scheduler was faster, but deep insurance-workflow coupling and data-ownership needs favored building on owned services with full control of PHI.
Governance & risk
API securityPHI handlingAudit loggingLeast privilegeCode review gates
Business outcomes & ROI / TCO

40% reduction in administrative overhead via automated appointment and insurance workflows; the platform reached 95% production-readiness. Staff hours redirected from manual entry to patient care translate to a modeled recurring OPEX saving and a fast payback on build cost.

User & citizen centricity

Patients self-serve appointments and insurance 24/7 instead of waiting on a phone line; the front desk handles exceptions, not data entry — a measurably better journey on both sides of the counter.

Case 04 · Data & Analytics · Prediction + Decision Support

Predictive No-Show Model & Market Intelligence Dashboard

85%no-show forecast accuracy
$500K+savings identified
Real-timestakeholder insight
EDW-backed
// Analytics & data viewBUSINESSClinic AccessMarket StrategyANALYTICSNo-Show ModelPower BITableauDATAEDWFeature StoreSQL MartsTECHNOLOGYPythonPandasscikit-learnETL
The challenge

Two decision problems, one data foundation: clinics were losing revenue to patient no-shows, and stakeholders lacked timely market intelligence. Both needed trustworthy data and the right modeling choice — not the most fashionable one.

My role

I built the data architecture and the analytics — the ETL into the EDW, the predictive no-show model, and the automated market-analysis dashboard surfacing insights in real time.

Approach & frameworks
  • ODS → EDW with governed ETL as the single source of truth.
  • scikit-learn predictive model on engineered features.
  • Power BI + Tableau for executive-facing dashboards.
  • Data governance & PII handling baked into the pipeline.
Key trade-offs & decision logic — why this, not that
DecisionConsideredSelectedWhy
Model Deep learning Classical ML (gradient-boosted / logistic)chosen A neural net was tempting, but on tabular clinic data it adds cost, opacity, and ops burden for marginal lift. Classical ML hit 85% accuracy with explainability clinicians could trust — the same discipline as choosing RAG over fine-tuning: pick the option that wins on cost, data fit, and maintainability, not novelty.
Latency Real-time scoring Batch (daily) scoringchosen No-show risk doesn't change minute-to-minute; daily batch scoring delivered the value at a fraction of the streaming infrastructure cost.
BI Build custom app Power BI / Tableau on the EDWchosen Reusing governed BI tools on the EDW beat a bespoke app on time-to-value and TCO, and kept one version of the truth.
Governance & risk
Data governancePII / PHI controlsModel monitoringSource-of-truth EDWAccess control
Business outcomes & ROI / TCO

85% no-show forecast accuracy enabled proactive outreach and reduced clinic revenue loss; the market dashboard surfaced $500K+ in identified cost-savings for stakeholders in real time. Choosing classical ML and batch scoring kept model-ops TCO low while delivering the business result.

User & citizen centricity

Fewer missed appointments means better access for the patients who need the slot, and leaders make calls on live data instead of last month's report.

Want a case study written up for your project?

I package engagements as executive one-pagers, architecture views, and short walkthroughs — exactly the multi-modal format decision-makers respond to.

Get in Touch → in · LinkedIn