Inspiration
Our Inspiration for this app was heavily drawn from our experiences as college students. We often find ourselves hunting people down to know what classes they are in and to find new classes for our next quarter. With the rise of LLMS and generative AI, we thought to outsource this work and ease the process for everyone: introducing Minz.
What it does
Minz allows students to connect alongside others to find the most optimal course schedules while still remaining in the loop with their friends. There are 4 main screens to our app. We have a signup/login in order to authenticate the user and make sure that they are who they say they are. Then we are greeted at the home page where we can see our own schedule and add courses that we are taking for the current quarter/semester. We can see a variety of different calendar views ranging from the day to the week to the month. There are 4 tabs at the bottom where we can redirect ourselves. The next one is the course recommendation page where one can ask the agi what is recommended to take for the next quarter and this is done using NLP and word embeddings so finding courses through an extensive catalog can be heavily simplified. Then we have our friends page where you can view what courses your friends are taking and also see what times those are to plan things to do together (before after or during class). Finally, we have an AI recommendation engine that will display a score between you and people in your direct classes based on what classes you are taking and the timings of those classes which will make it more enticing to approach them during class.
How we built it
Our application uses a multititude of technologies including a python flask server, Cohere AI's word embedding framework, Swift for the iOS infrastructure, UCSB's course catalog api for GOLD, and Google Firebase. We had a lot of fun learning about all of these technologies because they were very foreign to us, so it was a lot of trial and error and back and forth to get things working.
Challenges we ran into
One of the largest challenges we ran into was the word embeddings and storing it. We had a plethora of options to try and tackle this problem, but eventually settled on setting up our own python flask server so that we could take some load off of the phone in terms of memory and compute. This ended up working really well and it also served as our infrastructure for our AI recommendation algorithm which finds you the most optimal people to meet throughout a specific quarter.
Accomplishments that we're proud of
We are super proud of the entire suite and interface. This was definitely a large application with a lot of different features, but we believe that this is a very pratical application that can definitely be put into use by college students all around the world (if not high school and middle school as well). Leveraging the use of AI was definitely on our bucket list to start with so the fact that we were able to deliver on it was super fulfilling.
Log in or sign up for Devpost to join the conversation.