Bucky Boo

• 

  Preview of the frontpage.

Inspiration

• We bonded over the perils of scheduling season and vowed to create an enrollment experience tailored to students' interests.
• A dream, some Celsius, and a lucky fortune cookie (“Infinite patience produces ultimate reward”) later, we were on our way.

What it does

• With Bucky Boo, you can build a roster of classes based on what is hot and what is not.
• Think Tinder meets UW-Madison's Degree Auditing System: swipe on potential classes. If you match, you can chat with the very friendly AI chatbot who's thrilled to tell you all about each class.

How we built it

• Trained our own language model!
• Connected HTML, CSS, Python, and Javascript, and R for a user-friendly interface. Parsed a table out of very unusual PDFs in three different languages.

Challenges we ran into

Frontend:

• Connecting HTML, CSS, Python, and Javascript
• Node.js global failure
• First experience with all four languages (and maybe last) for frontend team (Cam and Zam)

Backend:

• First time ever working with language models (Michael)
• Parsing a table out of unusually formatted degree auditing PDFs in R, Java, and Python! (Shrey)
• API availability

All:

• Switching focus due to complexity constraints (originally Bucky Blanner)

Accomplishments that we're proud of

• Streamlined interface within a limited time
• The language model (duh)
• Learning multiple languages across 24 hours
• Surviving our energy drink consumption

What we learned

• The multifaceted operation of balancing frontend, backend, integration, and the extents of creativity
• How to restart when you've hit square one again (Bucky Blanner tragedy)

What's next for Bucky Boo

• Scrape RateMyProfessors API for extra functionality.
• If we make a bad match, we will implement a rating system so former and current students can weigh in.

Built With

• css
• flask
• html
• javascript
• openai
• os
• pandas
• python
• r
• requests
• selenium