KanYe.ai

Groq hosted, llama distilled, Deep Seek with advanced Kanye sandwich training. It's an AI tutor, that asks YOU to teach IT.

Comment

We made it. While we didn't try to make a societal impact within the 24 hours, if we get the chance to improve Kanye AI, we would very likely make it publicly available.

What it does

It's an AI which teaches you by first giving an explanation of a chosen topic, and then having you explain the concept back to it. It keeps track of the progress of your learning by evaluating how well you explained it.

How we built it

It's a web app with a ReactTS frontend and Python Flask backend. It reads user text input, and feeds it to a LLM using an API key from Groq. In addition, it features an in-house AI solution in the Plagiarizer.AI, which is admittedly not very good, but was built from the ground up.

Challenges we ran into

  • Getting the backend to sync with the frontend was a challenge throughout the Hackathon. A particularly large amount of difficulties came with our decision to use React.
  • It's native behavior is to reload the page upon submitting form data, constantly interrupting the backend
  • It has terribly unhelpful error messages
  • Using React with TS required using esoteric types with little to no documentation on them Unfortunately, our AI model was also very stupid:
  • It had a tendency to answer its own questions that it generated
  • It sometimes got sidetracked
  • It had a fixation with Kanye, hence the name of the project

Accomplishments that we're proud of

  • We built a mostly functional, and nice looking front end
  • We built two mostly functional, working chatbots that can teach a limited, but important, range of topics: How to do calculus, and How to make a sandwich.
  • We built an in house AI solution, Plagiarizer AI, which is not very good at all, but is entertaining nonetheless.

What we learned

  • We refined our fundamentals in using Python, and TypeScript
  • We experienced for the first time working in a mixed language codebase
  • We learned (the very very basics of) how to build and interact with AIs
  • We learned what NOT to do working in a mixed

What's next for KanYe.ai

  • Improving polish and features
  • Improving the Plagiarizer AI and replacing the Groq solution with Plagiarizer.
Share this project:

Updates