RAW

• 

  Asking the brain

• 

  Getting quizzed by the brain

Inspiration

Our teachers would give a lot of material and we'd take a lot of notes but during test days or if we needed to look for a certain fact, we'd have to search through all of our material and flip pages through our notebooks. If only there was a way to search quickly through everything and get quizzed on it.

What it does

Our project takes your notes and builds a brain based off of it. When you ask it a question, explain a topic, or recall a chapter, it will search through it's brain and give you an answer completely based on your notes and materials.

How we built it

We used embeddings to create the brain and used Ollama's llama3.2 for the LLM.

Claude Code was used to create our project

Challenges we ran into

LLM was generating false info (things not based on the notes we gave), libraries/packages version issues, LLM struggling with numbers (e.g. "Give me 5 questions on this topic" and returns a different number of questions)

Accomplishments that we're proud of

We're proud of using different principles of AI such as embeddings and local LLMs since this is our first time making an AI project with multiple layers

What we learned

We learned how to use git, how embeddings work, and training and integrating local models

What's next for RAW

Scale - Add more features, improve AI, expand use Consistency in responses - Responses should not vary in little things like number of questions, information, etc. Cleaner UI - Make the UI more intuitive to use and understand

Built With

• embeddings
• flask
• ollama
• python
• sentence-transformers