Product-minded engineer who ships end-to-end. I don't just build features — I track whether they work. 6+ years in production software development.

Based in Bogor, West Java, Indonesia. Building MSTORE — a multi-tenant POS platform for Southeast Asian SMEs. Engineering at BTPN Syariah (enterprise banking).

Available for remote contracts with US and global companies. Get in touch →

I build production systems end-to-end — backend services, admin dashboards, mobile apps, and the infrastructure that runs them. My focus is on multi-tenant SaaS architecture and Go backend development at scale.

Currently building

• MSTORE — Multi-tenant POS platform for Southeast Asian SMEs (Go/GORM/MySQL, Next.js, Flutter, Kubernetes)
• PAYSALES — Multi-tenant ERP platform for Global SMEs (Go/GORM/MySQL, Next.js, Flutter, Kubernetes)

Key projects

• 🏦 BTPN Syariah — Enterprise banking with Golang & Flutter
  • Sitepat — Customer-facing banking apps
  • Tepati — Loan management system
• ⛏️ PT Harmoni Panca Utama — Mining operations platform
  • HPU Digital Mining Platform — Fleet management, fuel monitoring & HSE for coal mining operations
• 🏢 Kano Solution — Enterprise solutions with React & Golang
  • IAG Cybera360 — Cybersecurity platform
  • Minerva — Mining Operation Petrosea

Currently — Senior Backend Developer @ BTPN Syariah + Building MSTORE

Previously

• Software Engineer @ Kano Solution (2020–2022)
• Internship @ EACIIT (2018)

• claudio — Anthropic SDK for the Dart ecosystem
• chainnusa — Southeast Asia's on-chain intelligence platform
• testcast — Opinionated Playwright TypeScript scaffolder
• geostack — Geospatial REST API boilerplate with Node.js, PostgreSQL/PostGIS & Docker

View all contributions →

• Multi-tenancy is a design decision, not a feature flag. Tenant isolation, RBAC scope, and cost-center modeling have to be designed into the schema, not bolted on later. I've seen both sides of this — first retrofitting multi-tenancy into BTPN Syariah's banking platform serving 4.1 million users, then building MSTORE and PAYSALES from scratch with multi-tenancy as a first-class constraint. The difference is night and day. Retrofitting into a single-tenant codebase is where most SaaS products quietly die — usually around the third enterprise customer, when someone asks “can we just add one more tenant?” and the schema says no. Getting it right early is expensive in design time. Getting it wrong is expensive in years.

• Contracts prevent drift — especially with AI agents. API specs, DB schemas, and service boundaries are the source of truth. Everything else is an implementation detail. I learned this the hard way building ML infrastructure at BTPN: OJK compliance meant every credit-scoring prediction had to be fully reconstructable — who requested it, when, with what input, what came out. Then again building RAG pipelines where the backend contract matters as much as the model itself. LLMs write code fast but respect exactly zero invariants — which is why contract-driven workflows matter more now, not less.

• If it's not observable, it's not shipped. The LGTM stack — Loki, Grafana, Tempo, VictoriaMetrics — is baseline on every project I own. But the real lesson came from a fuel anomaly detection system I built for mining operations: pure sensor data missed the theft patterns. It was the combination of GPS movement telemetry with fuel readings that made the anomalies visible. Three theft incidents caught in the first month, 8–12% annual fuel cost reduction. The takeaway: observability isn't just dashboards — it's knowing which signals actually matter and correlating across layers that weren't designed to talk to each other.

• Build for the worst-case scenario. The hardest engineering lessons I've had all share a pattern: the system worked fine in normal conditions. Then the VSAT link to a remote mine site dropped to 512kbps — or a self-intersecting polygon silently corrupted every downstream spatial query — or field agents in dead zones needed offline sync without losing loan disbursement data. At BTPN I cut data loss from 23% to zero with offline-first architecture. At Petrosea I built geofence breach detection that ran under 2 seconds against 50,000 spatial events per second. Happy-path engineering is easy. The edge cases are where systems earn their existence.

• AI-assisted, not AI-dependent. Claude Code, MCP servers, and semantic search are force multipliers — I use them aggressively. But I still read every diff, own every deployment, and stand behind every line that ships. The leverage transfers; the accountability doesn't. Tools accelerate execution, but they don't replace knowing why a design decision was made. If I can't explain the trade-off to another engineer at 11 PM during an incident, I shouldn't have shipped it.

Case studies — product & architecture writeups

• Sitepat — Customer-facing banking apps
• Tepati — Loan management system
• Harmoni Panca Utama — Mining operations platform
• IAG Cybera360 — Cybersecurity platform

Technical content on Dev.to, Medium, or Study Case

• Co-founder at Affan Tech — building tools for the developer community

More on LinkedIn · Resume · Study Case · Showcase Product

• 💼 Hiring? Available for remote senior/staff engineering roles. Email
• 🤝 Collaborating? Open to OSS contributions in Go, DevTools, or SaaS architecture.
• 💬 Just chatting? LinkedIn works best.

_{Based in Bogor, West Java, Indonesia · GMT+7 · 🕐 9 AM–5 PM PST / EST overlap for US team calls}