Inspiration
One of the hardest parts of being a CS major at Columbia is not the classes: it’s advising. The current system is overwhelming and unsustainable for CS advisors and students alike, and this reality motivated us to build CiCi.
What it does
CiCi is an AI chatbot trained on Columbia CS advising data. It uses this history along with its broader chatbot capability to answer any questions you have about the CS major, coursework, or career options at Columbia.
How we built it
Frontend: We initially wireframed our project in Figma. Then, we implemented it using the traditional JavaScript, HTML, and CSS.
Backend: We created embeddings using Columbia CS Advising documents and stored them in a vector database. We then used the OpenAI API to make a chatbot that would answer CS advising questions by searching through the vector database for the best answers.
Challenges we ran into
- Building responsive and logical UI
- System issues -- problems with Windows, downloading Python
- 0% hallucination from AI at this point is difficult
Accomplishments that we're proud of
- Building a fully working product in one day (none of us have really done this at a hackathon before)
- Successfully incorporating embeddings
What we learned
Technical Skills:
- How to use OpenAI API
- Deeplake database
- JavaScript/HTML/CSS, frontend responsiveness
Soft skills:
- Communication
- Confronting difficulties without getting overwhelmed
- Delegating work clearly between individuals
What's next for CiCi: the Columbia CS Advising Chatbot
We intend to present it to a Columbia CS professor who expressed interest in the idea later this week!
Log in or sign up for Devpost to join the conversation.