CodeRabbit
I joined CodeRabbit pre-seed as a founding product engineer and built the public foundations of the company end to end: the design system, the marketing site, the blog, and self-serve docs the whole team could publish to.
Visit CodeRabbit
Where it started
CodeRabbit is an AI code reviewer: it reads a pull request and leaves contextual, line-by-line feedback before a human opens the diff. I joined at pre-seed, when the product worked but almost nothing around it did. A developer landing on the site had no fast way to grasp what the tool actually did, and anyone outside engineering needed a ticket to publish a page.
A design system, not a pile of pages
Rather than designing screens one at a time, I built the system underneath them: type scale, color, spacing, and a component library the marketing site, the blog and the docs all draw from. A custom icon set came with it, so surfaces built months apart still read as one company.
Explaining the product without a deep-dive
A screenshot of a pull request thread means little to someone who has not sat through a slow review, and the audience was never only senior engineers. So the site explains the review flow visually rather than in prose: annotated diffs, worked examples, and illustration doing the job a paragraph could not.
Docs and a blog the whole team could publish to
The blog and the documentation run on the same system, with a self-serve authoring path behind them. Publishing stopped being an engineering ticket: anyone could ship a page and it came out looking like the rest of the site. Content was the main channel, and engineering time the scarcest thing in the building.
Written down, so it outlived me
The point of a system is that it keeps working once you leave. Color, type and the graphic language all ended up documented rather than living in one person's head, which is what let the brand hold together across a marketing site, a blog, docs and a product dashboard.