In this hackathon project, I developed the backend to support a personalized, AI-powered chatbot for engaging with educational media content. Starting with audio and video files stored in an AWS S3 bucket, I integrated AWS Transcribe to generate transcriptions and then leveraged OpenAI embeddings to convert those transcripts into vector embeddings. These vectors were stored in Pinecone, enabling efficient and relevant vector searches. When a user interacts with the chatbot, it retrieves the most pertinent video portions (along with timestamps) from Pinecone, sends this to OpenAI, and generates contextual responses tailored to the content.
Additionally, I implemented quiz generation on the backend, dynamically creating flashcards in question-and-answer format based on users’ learning styles. This allows the chatbot to not only answer questions but also reinforce learning through interactive quizzes on the frontend. I faced several challenges, including switching from DigitalOcean Serverless to AWS Lambda and Flask, ensuring that the backend was both scalable and high-performing. This project provided valuable experience in building AI-driven applications, working with a combination of AWS services and OpenAI, and developing a learning environment tailored to individual needs.
Log in or sign up for Devpost to join the conversation.