MD · MCS · PhD · CPHIMS · FISQua
Bridging clinical multiomics with scalable technology · HL7 FHIR interoperability · agentic AI · precision-medicine systems
I bridge two worlds that usually do not talk to each other: clinical medicine and scalable engineering. My work turns agentic AI, multiomics and hard interoperability standards into systems that hospitals can run on a Monday morning — not slideware that impresses on a Friday.
The turning point of that work was leading data standardisation for SATUSEHAT, Indonesia's national health interoperability platform. That is where FHIR R4, terminology mapping and master data management stopped being theory and became a national contract between thousands of facilities. Everything I have built since inherits that constraint: the payload must be right, in production, for people who will never read the spec.
🧠 Health AI · 🔗 Interoperability · 🧬 Multiomics · 🏥 Clinical platforms
|
Agentic assistants and decision support where the model proposes and deterministic rules decide. SmartClaim checks diagnosis, procedure and financing codes before a claim is submitted; IvanMD's CDS grounds suggestions in the encounter's SOAP; WISH turns a receipt photo into structured spending that three different assistants can reason over. Author of the open-source SATUSEHAT integration library — PHP, Node.js, Python and Go — plus SmarTerm, one terminology engine for ICD-10, ICD-9-CM, LOINC, SNOMED CT, KFA and KPTL. Running against the national platform, not a sandbox. Contributor to the national FHIR Indonesia Implementation Guide — the ID Core profiles (id-core-*) that SATUSEHAT payloads are validated against, built from FSH in CI. |
IDeRare — phenotype-driven exome analysis, published in JAMIA Open — and OMOP↔FHIR exchange for GWAS risk stratification. Genomic data is only useful when it reaches the clinician who is making the decision, in the format their system already speaks. EMR, LIS and ERP surfaces built for real hospital workflows: IvanMD, Clinix, SPECTRA. My clinical pathology practice — laboratory interpretation and whole-exome sequencing — is what keeps these architectures honest. |
| System | What it does | Link | |
|---|---|---|---|
| 🧠 | SmartClaim | Claim-integrity checks (ICD-10 / ICD-9-CM / INA-CBG) before submission | claim.ivanmd.id |
| 🔗 | SmarTerm | Shared terminology service: ICD-10, ICD-9-CM, LOINC, SNOMED CT, KFA, KPTL | terminology.ivanmd.id |
| 🇮🇩 | FHIR Indonesia IG | National ID Core profiles for SATUSEHAT, published from FSH via CI | ig.fhir-indonesia.org |
| 📦 | SATUSEHAT SDK | FHIR R4 builders + OAuth client + durable queue, four languages | PHP · Node · Python · Go |
| 🩺 | IvanMD | Medical record, teleconsultation and appointments (web + Android) | ivanmd.id |
| 🧾 | WISH | Receipt → structured expense, feeding three AI assistants | wish.ivai.app |
| 🧬 | IDeRare | Phenotype + exome pipeline for germline rare-disease diagnosis | repo · JAMIA Open |
| 🫁 | I3DR-Net | 3D lung-nodule detection on LIDC-IDRI with transfer-learning RetinaNet | repo · paper |
| 🏗️ | SPECTRA | Multitenant stock / ERP / POS / CRM — spec complete, build in progress | spectra.meditechku.com |
| 🔬 | Clinix | Clinical-pathology portfolio and laboratory reporting workflow | clinix.meditechku.com |
| 🔗 | IVAI Link | Invitation-only short links with QR and click analytics | link.ivai.app |
📚 The full list — 22 projects back to 2019, publications and experience — lives on ivanwilliamharsono.com. 🛰️ Every live instance, with its own status light, sits on my production pit stop.
Editor-time metrics stopped being honest the moment agents started writing code with me: hours fell while shipped work rose. This block therefore measures what landed — commits, pull requests, reviews, repositories touched — and whether the services behind them are answering right now.
Window: last 30 days · updated 2026-11-12 02:17 UTC
| Shipped | Count |
|---|---|
| Commits | 1562 |
| Pull requests opened | 129 |
| Pull requests merged | 119 |
| Reviews given | 0 |
| Active days | 30/30 |
| Private contributions | 0 (names withheld) |
| Repositories touched | 37 |
Where the work landed
Private — names withheld
private repo #1 279 ██████████████████
private repo #2 250 ████████████████░░
private repo #3 228 ███████████████░░░
private repo #4 78 █████░░░░░░░░░░░░░
private repo #5 76 █████░░░░░░░░░░░░░
Public
ivanwilliammd.github.io 23 ██████████████████
satusehat-integration 23 ██████████████████
iderare-pheno 10 ████████░░░░░░░░░░
satusehat-integration-node 8 ██████░░░░░░░░░░░░
satusehat-integration-python 8 ██████░░░░░░░░░░░░
30 private repositories · 1468 commits (names withheld)
Languages by bytes in those repositories
PHP ██████████████████ 11.26 MB
Vue █████░░░░░░░░░░░░░ 2.83 MB
Blade ████░░░░░░░░░░░░░░ 2.42 MB
TypeScript ██░░░░░░░░░░░░░░░░ 1.44 MB
Python ██░░░░░░░░░░░░░░░░ 1.02 MB
JavaScript █░░░░░░░░░░░░░░░░░ 0.76 MB
Live services (HTTP code · latency, checked now)
| Service | Code | Latency |
|---|---|---|
| Portfolio site | OK | 859 ms |
| IVAI | OK | 854 ms |
| IVAI Link | OK | 1113 ms |
| SmarTerm | OK | 829 ms |
| SmartClaim | OK | 824 ms |
| WISH | OK | 740 ms |
| Clinix | OK | 836 ms |
| SPECTRA | OK | 1202 ms |
| Pay | OK | 2175 ms |
| IvanMD Identity | OK | 1523 ms |
Measured from the GitHub API and real HTTP responses — what shipped, not how long an editor was open. Refreshed daily by .github/workflows/metrics.yml.
Six first-author papers — from a medical-school case report to a Q1 methods paper.
| Year | Paper | Venue |
|---|---|---|
| 2024 | IDeRare: phenotype + exome pipeline for germline rare disease · doi | JAMIA Open 7(2) · Q1 |
| 2024 | Phenotype and trio WES data in glycogen storage disease IV · doi | Data in Brief · Q3 |
| 2024 | Disease modelling in inborn errors of metabolism | Int. Medical Journal 31(6) |
| 2020 | Lung nodule detection with RetinaNet transfer learning · doi | J. King Saud Univ. · Q1 |
| 2019 | Lung nodule texture detection with 3D CNN · doi | CommIT 13(2) |
| 2017 | Immunohistochemistry in silent pituitary adenoma · doi | MEDICINUS 6(3) |
Stack, standards and conventions (click to expand)
Standards — HL7 FHIR R4, SATUSEHAT, ICD-10 / ICD-9-CM / INA-CBG, SNOMED CT, LOINC, KFA, KPTL, HPO, OMIM, ORPHA, OMOP, DICOM.
AI — agentic pipelines, RAG, prompt engineering, LangChain; scikit-learn, PyTorch/Keras for imaging and omics.
Backend — PHP 8 / Laravel, Python (FastAPI, Django, Streamlit), Node.js, Go; PostgreSQL, MariaDB, Redis; OIDC (Passport), Sanctum.
Frontend — Next.js, Vue 3 + Inertia, React, Tailwind, Vite.
Platform — Docker, GitHub Actions, Vercel, Cloudflare, aaPanel; zero-build deploys where the server may only git pull.
Conventions — code symbols in English, UI text in Indonesian; every change carries a test quoted before and after; a feature is done when it can be triggered through the API, not only through a dashboard.
The SDKs, the FHIR implementation guidance and the small utilities here are built in the open, most of them on unpaid time. If one of them saves you a week of integration work, you can sponsor the work — or send a change back. Bug reports, corrected profiles and better examples count just as much as code.
Advisory, architecture review, or hands-on delivery for health AI, FHIR/terminology infrastructure and clinical data platforms.
© 2019–2026 Ivan William Harsono · Founder of PT Meditech Konsulindo Utama · IVAI