Abstract
In modern technical and engineering pedagogical environments, undergraduate students are subjected to intense cognitive demands, particularly concentrated within brief mid-semester and end-semester evaluation cycles. A primary structural deficiency contributing to academic performance degradation, cognitive overload, and elevated anxiety is the profound reliance on highly fragmented, unstructured learning materials. SCHOLAR is engineered as a centralized, collaborative platform that functions as an advanced orchestration engine for academic synthesis, leveraging NLP, LLMs, and Retrieval-Augmented Generation (RAG) to autonomously ingest heterogeneous documents and compile an authoritative, syllabus-aligned "Perfect Note". The core system architecture consists of a mobile-optimized interface built with Kotlin and Jetpack Compose, connecting to a FastAPI backend. The platform uses LangChain for AI orchestration, implementing map-reduce pipelines for semantic deduplication and hierarchical merging of content. Retrieval is managed via a hybrid network utilizing cloud-native Pinecone and on-device ObjectBox vector databases, ensuring low-latency, context-specific query answering and real-time tutoring grounded strictly in the uploaded academic materials. Testing indicates substantial reductions in document synthesis processing delays, yielding a highly accurate, cohesive knowledge base that mitigates cognitive overloads and automates manual exam revision prep workflows.
