Koray Erkan

Koray Erkan

I’m a documentation engineer. I build documentation systems, not just documents: the structure, the pipeline, and the tools that keep the writing true to the software it describes.

What I do

I read the source, run the product, and write from what I find there, rather than waiting for an engineer to explain the feature to me. Then I build what the documentation needs to stay correct as the product changes: information architecture, docs-as-code pipelines, component libraries, and AI-assisted authoring tools with a hard rule that every generated sentence must trace back to a source.

Selected work

VirtualMetric DataStream documentation

The complete documentation for a security data-processing and routing platform: 550 pages, written and maintained by me alone. I also built the site itself (Docusaurus, React/MDX, a GitHub Actions pipeline to Cloudflare) and the Claude Code agents that kept one writer in step with several engineering teams.

DITA Toolset

A Claude Code agent bundle, in development, that generates DITA documentation from source code and then validates its own output. Its core rule: every generated element carries a warrant tracing it to code, a knowledge-base document, or a human decision. Anything without one fails the build.

Technical writing portfolio

Eight samples, one per document type: API documentation, installation guide, configuration guide, reference charts, release notes, tutorial, user assistance, and user manual.

Background

Before documentation, I spent two decades in software localization engineering for Microsoft, IBM, Google, SAP, Apple and others, which meant the technical end of the work: resource files, builds, validation tooling, and bug fixing. I write Python, C#, Go, PowerShell, and JavaScript, and I work in English and Turkish.

I’m based in Türkiye (UTC+3) and work remotely. I’m open to documentation engineering and technical writing roles. Get in touch.

Writing

The first posts are on their way. They'll cover documentation engineering, docs-as-code, and building AI-assisted writing tools that stay grounded in the source.