I am Umar Hadi, a Senior Platform Engineer based in Yogyakarta, Indonesia.

I work end-to-end across cloud, serverless, and bare metal — usually for small teams that own the path from database to deployment. My working line is short: good infrastructure should be boring to operate, cheap to run, and invisible to product teams.

The last few years have moved across DevOps, SRE, and platform engineering for a fairly varied set of products: a digital platform where I led a roughly 70% performance improvement and 50% infrastructure cost reduction, a multi-product software house running food, marketplace, and e-commerce verticals, a travel booking platform, an HRIS at real data scale, containerized AI deployments in restricted government environments, and most recently an AI agent platform.

In parallel I founded Zenvoye, a small software house focused on product development and platform engineering work.

Most of the systems I touch share a few traits: production traffic, real cost pressure, and operational habits that decide whether a system is still legible six months later.

I am interested in practical engineering more than novelty for its own sake. The questions I keep coming back to are:

  • What is the smallest architecture that can survive real use?
  • Which parts of a system need to be explicit because humans will operate them later?
  • How do we use AI tools without letting them blur ownership, source of truth, or judgment?
  • How do we write notes that make future work cheaper?

A few patterns keep showing up across that work:

  • A drift from cloud and serverless toward bare metal as systems mature and the cost or operational story justifies it.
  • Multi-vendor by default — AWS, Fly.io, DigitalOcean, bare metal — so design choices stay independent of any single platform.
  • Cost reduction by owning more of the stack: spot instances, lifecycle policies, mesh VPNs, self-hosted runners, rather than stretching managed services into shapes they were not meant for.
  • Boring operational choices over fashionable ones, when the team has to live with the result.

Outside contract work I run a small home cluster, experiment with personal AI agents and AI-maintained knowledge systems, and play more badminton than is reasonable for someone with a desk job.

This site is a public layer of my private notes. It is not a complete archive. It is a place for ideas that are safe, useful, and polished enough to share.

Good entry points:

Elsewhere: