CorAI

A better way to study for the 21st Century!

Comment

Inspiration

As college students, we noticed that taking notes during lectures can be really effective in the moment, but have diminishing returns later on, when reviewing for exams and preparing for later classes. Instead, we benefit most from application and testing our understanding. However, making review questions before fully grasping material is challenging. CorAI (Cor like Cornell notes) automates this process, by transcribing a recorded lecture and producing notes and review questions using Generative AI. Through this, we enable students to push their limits and learn more effectively than ever. We make CorAI available through corai.tech, where users can record lectures and receive generated review material.

How we built it

We host OpenAI's Whisper model to perform speech to text in real time and use GPT-3.5 to produce concise notes as well as review questions. The backend is hosted via Google Compute Engine, and the frontend is built with React and TypeScript.

Challenges we ran into

This was one of the first full stack projects we undertook by ourselves, so there was a lot to learn! We were using React for the first time, and handling the configuration difficulties of deployment throughout the hackathon.

Accomplishments that we're proud of

We're proud of putting together our first front end entirely by ourselves and that we finished the entire stack in the 24 hour time frame. We successfully hosted Whisper.cpp on our own servers to cut latency enormously compared to using an API. The fine grained control enabled us to make transcription much more practical for the end user.

What's next for CorAI

We'd like to expand on CorAI by scraping huge collections of questions and answers from the internet, to give our users more certainty in their studying. We'd like to add persistence so that users can organize their transcriptions, notes, and review questions, as well as a way to export them.

Built With

Share this project:

Updates