MASA
Multi-Modal Adaptive Study Agent — an intelligent AI assistant that transforms studying by digesting complex PDFs, generating quizzes, and tracking retention.
Overview
MASA (Multi-Modal Adaptive Study Agent) is a personalized autonomous tutor that parses complex multi-modal textbooks, diagrams, lecture slides, and video transcripts into interactive knowledge graphs and contextual flashcards with adaptive spaced-repetition testing.
Key Metrics & Performance
500+ pgs/min Document Ingestion
94.2% Retrieval Precision
3,500+ Active Students
Architecture & Design
Features a hybrid RAG pipeline using dense vector embeddings + BM25 keyword matching with cross-encoder re-ranking. The agent maintains a persistent learner knowledge graph in ChromaDB/PostgreSQL, diagnosing conceptual gaps and dynamically synthesizing personalized pedagogical explanations.
Key Capabilities
- Multi-modal PDF and diagram extraction with vision LLMs
- Adaptive quiz synthesis based on Bloom's taxonomy cognitive levels
- Spaced repetition scheduling based on SM-2 algorithmic memory decay
- Natural conversational Q&A grounded in verified source citations