Avothink
We're building a Bedrock-powered AI platform where teachers configure their personalized AI twin using specialized agents and pgvector.
Project Description
The project aims to both break down the communication barrier between students and teachers and modernize teaching methodologies.
The application currently includes a fully functional module that allows teachers to upload their study materials. These materials are deeply analyzed, embedded using Amazon Bedrock, and stored as vectors in the database, enabling advanced semantic search and follow-up queries.
Students can have either general conversations based on all available materials or focused discussions tied to a specific resource. The system automatically detects and adjusts the context of the conversation in real time, depending on its direction.
Additionally, the agent’s responses can be rendered as audio using ElevenLabs’ text-to-speech, making the learning experience more accessible and engaging.
https://github.com/linesbymartin/avothink-backend backend
https://github.com/linesbymartin/avothink-frontend frontend
https://www.canva.com/design/DAG7DMMmvOY/Q9QQyyfuV7IB0G8iIlQOyQ/edit presentation
https://www.figma.com/design/I9qgN4mL75mTQ5oMtYCVrT/Avothink?node-id=0-1&t=1zUUgubthi4rPfOp-1 figma design
Prior Work
The authentication module of the backend was build prior to the hackathon.