KnotEdu

Inspiration

We are facing a massive change in every aspect of life due to advancements in AI. One aspect that matters to everyone is education. With the world that we know shifting towards the unknown, there are many questions on how AI will be used in education. We believe teaching is one of the most novel professions, and trying to replace it with AI sounds unrealistic. But empowering teachers with AI sounds about right.

Teachers spend extra amounts of time outside of the classroom to ensure their students are receiving the education that they deserve. They prepare study guides, homework, fill out forms to parents whose child was not present that day, etc. They need to ensure that the student is continuing to learn outside of the classroom as much as they do it in the class, but there isn't any tool that is helping them at the level they need. They also have to improve themselves with precise evaluations and training materials, but it is impossible to have every lecture audited by another professional educator. This is why we built KnotEdu.

What it does

KnotEdu is an AI "Education Helper" that transforms lecture videos into a full learning ecosystem. It automatically generates transcriptions, chapter summaries, quizzes, external resource recommendations, and even provides delivery analytics for faculty. It allows the teachers to edit the generated results, giving them control over what they want to share with their students. Once the teacher makes the lecture public, the students can view the actual recording, summaries for chapters, questions for chapters, and extra resources to check out about that topic.

How we built it

We used Gemini models to empower our AI tools for transcription of the lectures, chapter/summary/question generation, and web search for extra resource recommendation. The Python backend downloads the YouTube (for MVP) video and passes it into Gemini API for AI related tasks, then we store the output on our website through MongoDB Atlas, and we provide the UI through Next.js and Node.js. We deployed everything on AWS EC2 instance.

Challenges we ran into

There has been many challenges building website, but there is an existent challenge we faced after deployment on EC2. The video downloading library we use in the backend (yt-dlp) is running into security issues, because API is considering the EC2 to be a bot. Due to limited time, we couldn't solve this issue, but the website is fully functional online other than video transcription. We are able to transcribe the videos on our local servers, and the online website is also being updated.

Accomplishments that we're proud of

We have been able to come up with an idea, design, innovation, and implementation in less than 36 hours. Our app is fully functional other than online video downloading, and we can solve that issue by adjusting code to use different libraries or since we can directly upload video recordings in actual product, that issue becomes irrelevant.

What we learned

We have been able to tackle all the issues, and implement all the features we wanted to implement. We used Google Cloud, used Gemini models for various custom AI tasks, deployed our app on AWS for the first time. It was a good learning experience and fun project that we were passionate about