SynapseLearn (Conversational AI Tutor)
A local-first, multi-agent conversational AI tutor designed for personalized learning. The system integrates a hybrid Retrieval-Augmented Generation (RAG) pipeline with adaptive memory and AI council-based reasoning to deliver context-aware explanations, student evaluation, and learning analytics. It features dynamic query routing (direct answer, retrieval, or web search), self-correction loops to reduce hallucinations, and multi-agent collaboration including tutor, pedagogy, evaluation, and memory agents. The platform supports both temporary and long-term memory modes, enabling personalized learning experiences while maintaining user privacy. The system combines local and cloud model execution (via Ollama and APIs), supports multimodal interaction (text and voice), and integrates browser-based context capture for real-time learning assistance. Compared to traditional RAG systems like GraphRAG, RAPTOR, and HippoRAG, this approach avoids heavy preprocessing by using a query-driven adaptive pipeline, resulting in lower computational cost, higher flexibility, and better scalability for real-world educational applications.
&image=https://ik.imagekit.io/miyasajid19/Portfolio/SynapseLearn.png&url=https://sajidmiya.tech){kind=link}
&image=static/images/portfolio-agent.webp&url=https://sajidmiya.tech){kind=link}
&image=static/images/ums.png&url=https://university-management-system-xvfp.onrender.com/){kind=link}
