Zain Haq
I write API documentation and developer guides, and I build tools that check documentation for accuracy. I've documented APIs, developer tools, and cloud platforms for more than seven years. I work docs-as-code, because documentation is part of the product and should ship with it. A growing share of documentation's readers are now AI agents that act on what the page says, so I write for them too.
Writing samples
These are published documents, recreated with fictional product names. You can read each sample in the browser, download it as a PDF, or fetch it as raw Markdown.
| Document | Type | Audience | Length |
|---|---|---|---|
| Authenticate to the Cove API How to generate API credentials, exchange them for OAuth 2.0 access tokens, and store both securely, with caching strategies and error references. | API authentication guide | Merchant and agency developers | 9 min |
| Cove Commerce accelerator for Meridian OMS Architecture, modules, and deployment of a pre-built integration between an enterprise order management system and a commerce and fulfillment platform. | Integration reference | Enterprise architects and OMS developers | 12 min |
| Update subresource integrity hash values for scripts in the Cove cartridge How to regenerate Subresource Integrity (SRI) hash values after changes to checkout cartridge files, so script tampering protections keep working. | Security how-to guide | Storefront developers | 3 min |
Featured project
AI-powered documentation quality scoring
AI helps teams produce more documentation, and more drafts mean more small errors for reviewers to catch. At Amazon, I built a VS Code extension that scores a documentation draft out of 5. The score combines editorial quality with technical accuracy checks against approved source material. A companion CLI scores web pages and Word documents in batches, so a team can measure quality across its entire library.
The tool gives reviewers a better starting point rather than replacing peer review. I calibrated it against documents I had already reviewed by hand and adjusted the scoring until its feedback matched what a strong reviewer would flag. To control costs, it reserves Claude for the checks that need judgment and leaves the rest to rule-based checks. I built it around the team's review process rather than my own habits.
- Form factor
- VS Code extension and CLI
- Model
- Claude Sonnet, used selectively
- Output
- Score out of 5, with the issues behind it
- Calibration
- Tuned against hand-reviewed documents
- Philosophy
- Human in the loop. AI assists review; writers decide.
Approach
Test first, write second
I test the product hands-on and read the source code when the spec is ambiguous. A claim doesn't go into the docs until I've seen it work.
Docs are code
Content lives in Git and ships through review and CI, like the product it documents. At H2O.ai, I owned the publishing pipeline: I led the docs migration to Docusaurus and connected the build to CI/CD. This site works the same way: every sample is a Markdown file that builds into the page you're reading.
Write for humans and machines
Documentation now has two audiences. When terminology stays consistent and headings stay exact, AI agents and retrieval systems surface the right answer for your users. Every sample here has a Markdown twin, and the site publishes llms.txt. The same structure makes content work with retrieval-augmented generation (RAG) and Model Context Protocol (MCP) servers.
Give AI a map
A model assembles what already exists. Deciding what's worth building still takes a writer who knows the product and the reader. When a model drafts for me, I expect most of it to be sound, verify it with automatic checks, and review the result the way I'd review a colleague's draft.
About
- Currently
- Technical Writer, Amazon
- Previously
- Senior Technical Writer, H2O.ai
- Education
- B.A. English Studies, Northern Illinois University
- Location
- Seattle, WA
I'm a technical writer with more than seven years of experience documenting APIs, developer tools, and cloud infrastructure. The job now has two sides, and I work both: writing the explanations that show how a product fits together, and building the systems that produce and check content with AI. At Amazon, I write documentation for Amazon Supply Chain Services and build tools that improve content quality across the team's docs, including the AI scoring tool featured above.
Before Amazon, I spent five years at H2O.ai documenting machine learning platforms and running the team's documentation platform. Owning a publishing system end to end, from Git to the published page, still shapes how I write.
Teams that work with me get direct questions and documentation feedback that engineers can act on. The goal is clear, accurate content organized around how people actually use the product.
Contact
Have an interesting documentation problem? I'd like to hear about it.
- zain.m.haq@gmail.com
- linkedin.com/in/zain-haq
- GitHub
- github.com/zainmhaq
- Resume
- PDF, one page