ClubScout | Devpost

Inspiration

As an incoming college student, it may be difficult to navigate through a new environment, especially if there are no resources that allow them to reach out to different communities. Finding a community is crucial to a student’s success - if a student surrounds themself with people who don’t have similar aspirations, they may not be able to see their full potential. This project was made with the purpose of connecting college students with clubs that suit their interests and personality the most; to provide them with a long-term community that can propel them towards networking and success, as well as life-long connections from meeting other people with the same interests. This would strengthen the likelihood of student success, further supporting the idea that club participation is an “essential need”.

Additionally, for scenarios where clubs that are “best fit” for a user are sparse, we provide advice and resources on how to approach making a club that is more tailored to the user’s interests. This highlights the success aspect of a college student, as it not only encourages them to create a club to promote leadership and individuality but also provides another inclusive community for those who don’t have a club best fit for them.

What it does

The user answers a series of questions that range from their academic to personal interests and with the use of OpenAI’s API and data from Hunter College’s official student club website. The website outputs which clubs are the best fit for the user based on the questions previously answered and explains why they are best-suited for the user.

How we built it

We first built a custom web scraper to obtain the data necessary(club information) from the Hunter College website into our Firebase database. The club information that was important was the club name, club type, the description, and an assortment of media links. The front-end was designed on Figma and then built with Next.js, React, TypeScript, and Tailwind CSS. Populating our vector database was done in Python using Pinecone as the vector database and Openai for the embeddings and recommendation system.

Challenges we ran into

Our initial webscraper was coming back with too much information and oftentimes duplicate results - returning over 4,000 clubs instead of the ~100 clubs at Hunter Difficulty populating the Firebase database and pulling the correct information for the UserProfile, ClubProfile, and Results/Recommendation Page

Accomplishments that we're proud of

Each single teammate learned a new technology, further contributing to their skillset. Shohruz learned about how to make RAG-assistant using vector database embeddings through Pinecone and combining that with. OpenAI. Afnan learned how to web scrape in python using beautiful-soup. Yna and Nick both learned Next.js and TypeScript. One of our proudest moments was seeing the website come to life and results populate both in the interface and the database. Getting to see everything come together is one of the best feelings in the world! We are also proud of the fact that our team has gotten to know one another much better throughout this Hackathon and was able to collaborate nicely to create this project.

What we learned

We learned that creating a product for university is not easy! There are so many “CUNY Essentials” to choose from! There was so many possibilities and so many features that we wanted to add. In order to create something presentable by the end of the hackathon, we needed to focus only on the core components. Through this experience, we can say that there is always opportunity for us to contribute to the betterment of the CUNY experience. Never stop!

What's next for ClubScout

As of right now, we created the base product for Hunter College Students. The goal would be to create a CUNY wide version of our product to ensure that students from anywhere are able to join the clubs within their university or draw inspiration from another CUNY’s club and create one of their own.

We would like to continue to build additional features for both users and clubs such as creating an interactive calendar so that clubs can add their events there and users are able to view this events. This would help the product become a 1 stop shop for all things clubs. Another feature that would help would be able to have anonymous reviews of clubs so that others can get a sense of what the club is about!

Built With

beautiful-soup
clerk
next.js
openai
pinecone
python
react.js
tailwindcss
typescript

Submitted to

CTP Hacks 2024

Created by

I worked on the backend and polished the frontend. I populated Pinecone's vector database with the club data provided by Afnan's web scraper. I created the form to create the recommendations and implemented OpenAI's API to generate recommendations based off of the similarities between the form inputs and the existing knowledge base of embeddings in pinecone.

Shohruz Ernazarov
I worked on creating the web scraper to grab information from the hunter clubs page. The same practice can be used on other club pages for other schools, thus the extent to which our application could grow is endless. I also contributed to connecting links and creating the recommendations page, along with the user profile(unfinished) using the the UI design created by Nick and Yna.

Afnan Ebrahim
I worked on the design for our website as well as the backend for fetching and displaying the clubs found using the web scraper that Afnan created. I used Firebase to add the web scraped data of all of the Hunter College clubs into a collection, and fetched that data that shows in our “club directory” page.

Ynalois Pangilinan
I worked on helping design the UI for the website using Figma and working on implementing some of the pages for front-end. Got to learn some new languages and skills!

Nicklaus Yao