What it is
This site started as an Astro Paper setup, then gradually turned into a personal system for publishing ML writing, LLM workflow experiments, and project pages.
The goal is not to build a complicated CMS. The goal is to keep publishing simple while making room for long-form technical writing, experiments, and project documentation.
What changed
- Reworked the homepage and information architecture to remove most of the template feel.
- Added a dedicated projects collection so project pages do not live in the same content bucket as long-form writing.
- Migrated historical machine learning writing into a consistent frontmatter format.
- Switched math rendering to compile-time KaTeX for archived ML notes and formulas.
- Deployed the site through GitHub and Vercel with sitemap, Open Graph, and search indexing in place.
Why it matters
I wanted a publishing workflow that stays close to code: Markdown files, git history, predictable builds, and no hidden admin panel.
This project is also where I test how a technical blog should feel when it is used as both a writing archive and a lightweight surface for experiments.
Next steps
- Add more real ML/AI project entries and improve the visual identity.
- Design a custom Open Graph image instead of relying on the default one.
- Keep refining the balance between writing, projects, and long-term maintainability.