Inspiration & What it does

After undergrad, every member in our group hopes to go to graduate school in pursuit of a Ph.D.! However, this process is oftentimes tricky and more complicated than it needs to be. Our project originally aimed to simplify this process by using a web scraper and a large language model to match a user to a variety of institutions with research labs that may be of interest to them. However, given the constraints of a Hackathon, we simplified the project to use the same tools to generate labs within the UMBC Computer Science department that a student may be interested in.

Challenges I ran into

One of the many challenges we ran into was the definition of scope. We had the initial idea of a Grad School Web Scraper that worked for every university across the U.S. because of schools' confusing ways of communicating their graduate resources (which is why our idea is a necessity!). During the competition, due to the limited time and resources we had available, we simplified the idea to a UMBC Lab Finder that can be built up into a fully-fledged Graduate School Information Finder later.

Another issue was choosing the right LLM, between local llama, open ai gpt, gemini and groq was difficult. Additionally, as a first-time hacker who has never done web development, I was learning on the fly while writing code.

How I built it

We integrated a data pipeline with an AI-powered searching algorithm to create an easy-to-interact searchable data set. Used LLMs to help create the data set, and Elastic Search to search through it

Accomplishments that I'm proud of

Got the AI-powered searching algorithm up and running. This wasn’t trivial, it required an intelligent setup for turning our raw data into an indexed, searchable data set, and an equally technical way to generate a query from raw user input

What I learned

SOO much, got into LLMs and semantic searching for the first time, learned about creating a backend with Mongo, and learned how to be a good hacker in 24 hours. I have never been more locked in HTML, CSS, LLM, Web Scraping :D!

What's next for LabFinder

Given how applicable this project is to the members of our group, let alone the greater academic community, we all strongly hope to take this project and push it further into what it was originally intended to be: a fully realized platform to make the graduate school search less grueling than it currently is.

Built With

Share this project:

Updates