HARUMI SPACE — Ideas and work
01 · Threshold
A starting point, not a fixed direction
02 · About
The thinking and experience behind the work
About me
Platform / Backend Engineer
I am Harumi Morikawa, a platform and backend engineer based in Japan. I build, deploy and operate systems end to end: enterprise Java, Camunda workflow systems, Kubernetes and GitOps.
I care about what happens after a design leaves the drawing board: how it is deployed, observed and maintained. Here I share both the things I build and the reasoning behind them.
03 · Selected work
From an idea to something that actually works
Running the cluster this whole lab sits on
One delivery path and one internal naming scheme across the registered applications; they discover themselves from their own repositories and retire the same way. Public identity policy is explicit, and existing metrics raise alerts instead of sitting unread.
Bare-metal Kubernetes infrastructure and an Oracle Cloud cluster in London host the lab workloads. Applications are discovered from repositories rather than registered by hand, then take the same route: test, build, scan and a separate deployed-tag commit.
Running the full Camunda stack
Identified connection reuse — not connection age — as the cause, and removed an unverified policy instead of layering another tuning fix on top of it.
Operating a full Camunda 8 orchestration stack (Zeebe, Operate, Tasklist, Optimize, Web Modeler, Keycloak) behind one unified public domain, and diagnosing a connection-handling "fix" that was quietly making things worse.
Harumi Platform — retiring three admin tools into one
Consolidated three separate operational surfaces into one controlled console, without discarding the data that lived in the tools being retired.
A single console for cluster health, resources, docs and connection info — replacing three separate admin tools that had each accumulated a different slice of the same operational picture.
LexTrace
A learning platform for technical language
A self-hosted study platform built around your own study routes: import material in a fixed format and build custom routes from it. The code is now open source (MIT) and the hosted service is being wound down.
04 · Ideas
Ideas worth testing in the real world
HotThere
Real places. Real-time vibes.
An iOS-first, map-first app for what is happening at a place right now, with explicit trust (verified, stale or reported) and explicit provenance (community or official).
Kashiwa Yui (kashiwa.help)
A Japanese-first civic MVP
A public civic-governance MVP for the city of Kashiwa. It publishes what the service does with people's information and which Japanese rules were judged to apply, in Japanese, English and Simplified Chinese.
maintenance-diorama
A maintenance page that is a film still
A cinematic maintenance page served from a Cloudflare Worker: a toon-shaded three.js diorama of a rainy-night Japanese street-corner construction site, with a live clock, the expected end time and a strip of the services the platform runs.
05 · Research
How decisions become systems
Library — notes and research
The long-form reading
Where the long pieces are kept: how the platform is laid out, how a change travels from a commit to production, and the dated history of the work.
Working paper — when my checks answered the wrong question
Eleven operating episodes, written entirely by AI assistants from their own practice
A successful check, a remembered instruction or an approval route meant less than I assumed. The paper follows eleven episodes of running this platform: verification against an earlier run, gates that failed silently, a recovery that needed the broken system, and a reboot that hung. Every word was written by AI assistants (Claude and Codex) from the records they kept; it is a working paper, not peer reviewed.
Camunda research — a role that closes itself
Designed and ran a small Camunda 8 experiment: a role opened by a message, three parallel lanes whose work arrives as real child process instances, and a role that ends after its deadline once no child is still active.
The first experiment in a small collection about process design and runtime behaviour in Camunda 8. A role opens on a message, runs for a window, and drains: work that arrives during the window becomes child instances whose parent and root are the role, so the number of root instances stays small. A tiny operator page opens, feeds and cancels roles and edits a monthly or interval schedule.
06 · Systems
The infrastructure behind these projects
Platform — live and in operation
A public summary of the systems that run this site
This Space traces the path from source code to a running platform. A small public summary shows available health counts for applications, nodes, pods and disk checks, directly from the platform. There is no simulated dashboard.
The public view is intentionally limited to aggregate figures. It does not expose hostnames, addresses, network topology or credentials. For the architecture and workflow design behind the numbers, follow the links below.
07 · Next
Still building. Still curious.
There's always another chapter.
What comes next
The work does not end with a launch. I want to keep building useful things, operating them, writing down what I learn and sharing more of the process. This space will change as the work changes.
Find the source on GitHub, the story in Milestones, or get in touch. The next chapter is open.