Technical writing, documentation architecture, and accessibility QA. I write the content people reach for when something isn't working such as support documentation, alt text, and reader-assist copy. And I test that it actually works for the people using it.
Owned support documentation and its information architecture for a civic technology platform serving 9,000+ organizations across 112 countries.
The work was structural as much as editorial: restructuring the help center so customers could self-serve the highest-volume questions, standardizing terminology, and maintaining the style guide that governed tone and usage across marketing and product.
That structure paid off three times. A well-labeled, consistently-tagged documentation set was needed to prepare Intercom's FinAI to ingest the data and transition to partially machine-assisted support for NationBuilder. It's also the thing answer engines can parse. When I later built the AEO practice, the docs were already legible to retrieval systems because the information architecture was sound.
Wrote alt text and applied web accessibility practices across the help center and marketing pages, then QA'd website content in Figma before it shipped in HubSpot, catching structure and copy problems at design stage rather than after publication.
Also ran QA and software testing across product releases, so documentation shipped accurate on day one rather than being corrected after customers found the gaps.
ADA-compliant alt text and reader-assist copy for New York City Department of Education textbooks and accessibility devices, plus technical writing and proofing for NYC government, Chamber of Commerce, and Department of Health materials.
Educational alt text is a specific discipline: a diagram in a textbook has to convey the same information to a screen reader user that a sighted student gets from the image, without editorializing and without becoming so long it's unusable.
Authored AI guardrails documentation for a coaching product, defining boundaries against unauthorized practice risk under emerging state AI regulation, and specifying prohibited outputs to keep the product a supportive tool rather than a stand-in for licensed care.