Kyooro
Ethereum-based AI tutoring platform using RAG for personalized learning, featuring MetaMask/Web3Auth login, crypto payments, and on-chain agent access keys.
Motive
With the rise of AI in education, there was a critical need for a system that restricts AI responses strictly to teacher-provided materials to prevent hallucinations, while simultaneously using blockchain technology to secure payments and manage access rights for content creators.
Overview
Kyooro is a decentralized AI tutoring platform that utilizes Retrieval-Augmented Generation (RAG) to create subject-specific AI agents[cite: 26, 36]. [cite_start]It integrates Ethereum-based smart contracts to manage course enrollments and 'Agent Keys,' ensuring that only verified students who have paid via crypto can access the proprietary training data uploaded by teachers
User Flow
User authenticates via Web3Auth or MetaMask SDK [cite: 9, 23] [cite_start]-> Teacher uploads PDF notes which are vectorized and stored in ChromaDB [cite: 40, 46] [cite_start]-> Student purchases course access on the Ethereum blockchain -> Smart Contract records the 'Agent Key' -> Student queries the AI, which validates on-chain access before retrieving context-specific answers
Future Plans
- Implement Live Virtual Classrooms using video-conferencing APIs (like Agora)
- Develop AI-powered student insight dashboards to track performance and attendance
- Add automated homework grading and submission systems
- Implementing a Retrieval-Augmented Generation (RAG) pipeline to restrict AI responses to specific teacher-uploaded datasets
- Integrating decentralized authentication using MetaMask SDK and Web3Auth to bridge traditional Web2 login flows with Web3 wallet generation.
- Architecting a hybrid system that verifies on-chain 'Agent Keys' (Ethereum) before granting access to off-chain vector databases (PineCone)
- Designing a 'Key creation engine' that securely maps student blockchain transactions to specific classroom content permissions in the vector database
- Managing the latency between Ethereum block confirmations and the immediate availability of the AI agent for the student.
- Ensuring the AI agent strictly cites sources from uploaded PDFs while preventing hallucinations from external data