youreindebt

Inspiration

Ontario's 2026 OSAP changes cut the grant portion of aid from 85% to 15%, converting thousands of dollars students relied on into loans overnight. We wanted to make the impact of that change impossible to ignore, and give students a way to fight back.

What it does

youreindebt is a web app and Chrome extension that shows Ontario students exactly how much the OSAP cuts cost them personally, then finds live scholarship matches to close that funding gap.

The web app walks you through a 5-step profile wizard, calculates your OSAP before and after 2026, and crawls live scholarship sources in real time to surface funding you're actually eligible for, matched by school, program, income, and background.

The Chrome extension surfaces your projected debt and matched scholarships while you browse school and program pages, so you never make a decision without knowing what it'll cost you.

How we built it

Frontend: React, TypeScript, Tailwind CSS
AI matching: Google Gemini 1.5 Flash for scholarship filtering and eligibility reasoning
Embeddings: OpenAI text-embedding-3-small via OpenRouter for vector similarity search
Web crawling: Firecrawl for live scholarship and tuition data
Database: Supabase with pgvector for semantic scholarship search
Chrome extension: Chrome Storage API, content scripts

Challenges we ran into

Getting the embedding pipeline consistent was the hardest part, we had dimension mismatches between ingestion and query time, and the Gemini SDK defaulted to API versions that didn't support our chosen models. We ended up switching to direct REST calls to fix it. Merging two separate codebases mid-hackathon while keeping both the web app and extension functional was also a real challenge.

Accomplishments that we're proud of

A single form that outputs a real, personalized dollar figure, not an estimate range, not a generic calculator, but your specific funding gap based on your actual school, program, income, and living situation. That number hitting the screen is the moment the product works.

What we learned

That the real complexity in AI-powered apps isn't the AI, it's the data pipeline. Getting clean, consistent embeddings across ingest and query, handling live web scraping gracefully, and making sure the filtering is strict enough to be trustworthy took far longer than building the UI.