Shirabase

The job I thought I had

I started out meaning to redesign my university's course management system. It was clunky, the information architecture fought the user, and a redesign was the obvious deliverable. That framing lasted exactly as long as it took me to ask someone else about it.

What the feedback changed

I talked to students and professors instead of going straight to Figma. What came back was not "this interface is hard to use." It was that the system holds information nobody gets value from. Syllabi are written once, filed, and never read again; even though they already describe, in detail, what a student will be able to do by the end of a course.

So I stopped redesigning the container and started working on what was inside it. The question became: can the syllabus that already exists be turned into something a student can put on a job application?

How it works

Shirabase takes unstructured syllabus text and extracts the concrete, market-readable skills buried in it, then presents them as something actionable rather than administrative.

The analysis runs locally in the browser. That was a deliberate constraint: course material and a student's academic record are not data I want to hold, and a tool that uploads nothing is a tool that is much easier for an institution to say yes to.

It is deployed on Vercel with analytics wired in, so I can see what people actually do with it rather than guessing.

What I took from it

The useful lesson was not technical. The brief I was given described a symptom, and the redesign everyone expected would have shipped a prettier version of a thing nobody needed. Asking first cost me a few conversations and changed the entire project.