Intellex AI

Workspaces
Login
Flash card
Answer evaluation
Quiz

Inspiration: Dense verbose readings

We were met with having to read extremely dense content and thought that it was a ridicously inefficient way to learn, especially since students have so little time nowadays and classes demand 10-15 hours each every week. So that's when we decided to make an app that could help make studying easier, and help you spend less time making study materials, and more time doing things that count and actually help.

What it does:

It makes quizzes--with both MCQs and open-ended questions. It answers any question--based on the documents provided-- and most importantly, if the content isn't available in the document, it actually says so. No more hallucinations while you're trying to study, and stuff that's outside of the course, won't impact your learning.

How we built it

The goal was to create a tool where students could upload their study materials (like PDFs and notes) and get a personal AI tutor that could answer questions and generate quizzes based on that material. We built the frontend with Next.js and React, and it uses Zustand to remember who is logged in. The backend is a Python API that handles user authentication, document uploads, and chat requests. It uses LangChain to process and store document information in Pinecone, a cloud-based vector database. This system runs a "Corrective RAG" workflow, which generates answers, grades them for accuracy, and uses web searches to self-correct before sending a final response to the user.

Challenges we ran into

We had some trouble with getting the APIs work properly. The code of accepting a document upload was inconvieient to work with, and the quality testing for the AI answers were also difficult to code.

Accomplishments that we're proud of

We're proud to have a fully functioning prototype for Intellex AI, and we're very proud that our AI tells the user when the answers they're expecting actually doesn't exist within the documents provided. AI hallucinations were a big problem with other apps we use.