Syllabot

Inspiration

In hearing the testimonials of professors who spoke at the opening ceremony, we realized that there was a significant need in the market for a centralized resource sharing platform amongst university professors. Leveraging a solution to this issue with AI was an innovative idea that our team wanted to pursue; the tool we came up with, as a result, was one that could be applied to expanding curriculums regarding climate and sustainability, but also to a variety of other issues pertaining to any given user’s values as well.

What it does

A professor first creates an account with their name and institution email. Once they log in with their account, they gain access to a portal containing a variety of tools aimed at enriching their existing curriculums. Our main tool, an AI similarity model, takes in a user input and a course description of the professor’s class before comparing it with that of hundreds of other courses across the country, allowing professors to find inspiration from the curriculums of other leaders in the fields.

Still confused? Here’s an example!

Let’s say a finance professor at Columbia was looking to expand their curriculum to include sustainability and its effects on the stock market. Looking for inspiration on how to go about this, they click into our AI tool and type “I want to incorporate sustainability into my curriculum with emphasis on how it affects the stock market”.

The professor doesn’t have to specify what his class is about- the course description is already saved and linked to their identity! Our AI makes comparisons between their course and hundreds of curriculums from schools near and far, taking into account the professor’s want for sustainability and its effects on the stock market. After a few moments, the professor is able to see the top ten courses most similar to their course + the input they provided; from there, they can click into any of these other courses for their full syllabi, opening up the possibility for them to contact other professors (not necessarily from their school!) for additional help in expanding their curriculum.

How we built it

The backend was built entirely in Python3. We used SQLite to store account data, and SQLAlchemy for course data (which we manually collected). Data analysis was also done in Python and utilized the API of BERT, Google’s language model, in combination with the data we fed to it. The frontend is built with HTML/CSS and connected to the backend with Flask. Designs and flow models were created on Figma.

Challenges we ran into

Utilizing an AI to make comparisons on specific data is… hard. The backend team spent an immense amount of time attempting to incorporate the model with many forms of data. Translating Figma designs into frontend was also quite difficult. Passing along user inputs to the backend for calculation, then results back to the frontend to display was also an arduous task.

Accomplishments that we're proud of

What our project resulted in is more than just a “CTRL-F”- it’s a genuine AI that scales and gets better as more people join and add their own course lectures and notes to it. We’re proud because this project has an incredibl3 capacity for growth beyond this hackathon.

What we learned

It was many of our members first time working directly with a machine learning model, so getting first hand experience in building one for a specific purpose was really cool for all of us. It was also half the team’s first hackathon, and everyone’s first hackathon working with such a large and diverse group (engineers had only previously worked with engineers, as did designers, etc) and working in a lively group such as this one was a great collaboration exercise.

What's next for Syllabot

Syllabot has so much room to grow! Our immediate next steps would be to incorporate course material editing functionality, so that professors could change the materials they’ve uploaded, indicate that they teach multiple classes, etc. We would also want to expand our dataset- by the end of the hackathon, we had more than 200 course data points across four different schools, but Syllabot could become incredibly powerful as it gains more data to make comparisons with. Features such as professor to professor messaging capabilities and a mobile app were also ideas that were discussed when looking at the future of Syllabot.