ScholarAI

Inspiration

During my senior year of high school, I was only limited to applying to in-state colleges since my parents couldn't afford for me to go to nicer colleges out of state. I looked for a lot of scholarships to try and help my parents but it was difficult to find ones that fit me perfectly.

I made this app so future high school seniors will have an easier time finding scholarships to help afford to go to their dream college

What it does

The app takes in a natural language prompt and searches scraped data from the web in a vector database

How we built it

I scraped data from collegescholarships.org and turned them into embeddings which I stored in the vector database. The frontend consumes the natural language input and turns it into embeddings to do a vector similarity search using an rpc call

Challenges we ran into

Scraping the website and learning how to clean a dataset to process into embeddings properly

Accomplishments that we're proud of

Learning how to clean up data, modeling the data to include embeddings, and learning how to do vector similarity searches

What we learned

How embeddings work and how to take user input and use them as embeddings to search the database

What's next for ScholarAI

Scraping more websites to have a larger database to pull from and to improve the UI/UX

Built With

• next.js
• pgvector
• rpc
• supabase
• tailwind
• typescript