Taking the longest route: a semantic course matcher
I've noticed something about myself: whenever a problem has a simple solution, I somehow end up choosing the most complicated path possible.
Applying for Erasmus means matching courses at your home university with courses at the partner university — digging through catalogs, reading descriptions, checking ECTS credits.
Instead of doing that by hand, I started building something: upload your syllabus, choose the partner university, and the system crawls its course catalog, extracts descriptions and credits, converts everything into embeddings and runs a semantic similarity search to find the closest matches.
- A crawler to collect course pages
- A parser to extract structured course content
- Vector similarity to compare topics between courses
Definitely the longest route to solve a small problem. But sometimes the most interesting tools come from refusing to take the easy path.
