This portfolio
The site you are reading: a static, dependency-light build with its own automated image pipeline and a content layer that renders to HTML, structured data and machine-readable text.
Role
Designer & Developer
Duration
Ongoing
Year
2026
Status
Live
0
Runtime dependencies for images
0
Secrets needed to build
Overview
A portfolio is a small project, which makes it a good place to be strict. This one is built to demonstrate the same decisions I would make on client work: own your content, own your assets, and avoid dependencies you cannot justify.
The problem
A portfolio has to do two jobs that pull in opposite directions: convince a person to start a conversation, and be legible to the search engines and AI assistants that increasingly answer 'who should I hire' before a person ever visits. Most portfolios do the first and ignore the second, or chase the second and become unreadable.
Constraints
- Static hosting with no server, no API routes and no image optimizer
- Content must be maintainable years from now without re-learning the codebase
- No dependency that cannot justify its weight
- Must render correctly without JavaScript
What I built
Everything renderable lives in one typed content layer. The same objects produce the HTML a person reads, the structured data search engines consume, and the plain-text representations AI systems can parse. Those three cannot drift apart, because there is only one source. Images are handled by a scripted pipeline: an authored manifest records which images exist and what they show, an idempotent upload script pushes them to a CDN and writes back dimensions and placeholders, and the generated map is committed so builds need no credentials and no network. Because the site is a static export, there is no image optimizer, so images are served through explicitly requested formats and a responsive source set rather than a framework component.
How it fits together
- Next.js App Router exported to static HTML, with no server and no runtime rendering
- Typed content layer that renders to HTML, JSON-LD and plain-text representations from one source
- Cloudinary delivery through hand-built URLs, with AVIF and WebP offered explicitly and the browser choosing
- Image dimensions recorded at upload time, so every image reserves its space before it loads and the layout never shifts
- Blog written in Markdown, processed at build time
Running in production
Served by Vercel as a static export. The build needs no secrets: the asset map is committed, so a checkout and an install are enough to reproduce it.
My role
- Design, content and build
- The image pipeline and its upload tooling
- Structured data, metadata and machine-readable output
- Accessibility and performance
Outcome
- Adding a project means editing one typed object. The page, the structured data and the machine-readable version all follow
- Images are automated end to end: no manual uploading, no hand-written URLs, no forgotten alt text
- The site works without JavaScript and reserves space for every image before it loads
Last updated
What I delivered
One typed content layer renders HTML, structured data and machine-readable text
Scripted, idempotent image pipeline, with no manual uploads and dimensions recorded automatically
AVIF and WebP offered explicitly, after measuring that automatic format selection was not serving AVIF
Zero layout shift: every image reserves its space before it loads
Renders without JavaScript