AI Scheduler | Devpost

Inspiration

All four team members of this project are students of Georgia Tech, and we've all racked our heads trying to figure out what classes we need to take next semester and how to best integrate them into our existing extracurricular and personal commitments. With the mainstream adoption of ChatGPT as everyone's favorite personal assistant, our thoughts went to how LLMs could be applied to solving this problem for us, giving each student at Georgia Tech—and eventually every college campus—the best schedule for themselves with just a few clicks and prompts.

What it does

Backend: AI Scheduler collects data from GT's OSCAR registration system, ratemyprofessors.com, and GT Course critique to inform an LLM model that backs the chatbot students interact with on our website. Since our website has users log in at the start of their session, schedule data is saved so they can come back to it.

Frontend: The user can interact with a chatbot to see the details of various courses and add them to a calendar on the website to build their schedule.

How we built it

We built it using:

LangChain
Python
Google Firebase
React
GT Course Data from Oscar, RateMyProfessor, and Course Critique

Challenges we ran into

Deciding on what GT Websites and course data should be used (We searched over 14k websites) Learning how to use the OpenAI Azure API Python crashed when we gave it too much data in a script Reading the data and cutting repeat data from RateMyProfessor since the format of the names was not consistent (First Middle Last vs. Last, First Middle)

Accomplishments that we're proud of

Learning how to Prompt GPT-4 Understanding and utilizing LangChain for embeddings and agents Used React and Tailwind for the first time and created the front end of the chat feature from scratch (without a style sheet or imported library)