Scatch: finding the scholarships a student actually qualifies for
Scholarship calls in Bosnia and Herzegovina are long, written in different formats and hard to check against your own situation. Scatch does that check for the student.
The problem
Students have to read dozens of calls to find out whether they meet the conditions. Eligibility rules hide in long paragraphs, and deadlines are easy to miss.
What we built
Scatch collects scholarship calls from public sources and uses an AI model to turn each call into structured eligibility criteria. The student answers a few questions, and the results are sorted into "for you", "maybe" or hidden, each with an explanation of why.
- Matching runs in the browser, so the profile stays on the student’s device until they sign in
- An application guide for each scholarship, with quotes from the original call
- Optional email alerts for new matching calls
- Optional CV import and Google sign-in
Under the hood
The front end is Next.js. A Python scraper collects the calls on a schedule and an AI model extracts the criteria, which a person can review. Data is stored in Firebase (Firestore) in the EU, and the matching itself is deterministic so results are consistent and explainable.
Why it matters for your project
Scatch shows how we use AI where it helps (reading messy documents) and plain code where it must be predictable (matching and ranking). That split is how we approach AI work for clients too.