Offline Biomedical Knowledge Engine for Evidence-Grounded Summarization Using Medingest AI

Abstract

The rapid growth of biomedical research data has made it difficult for researchers and clinicians to quickly access reliable and relevant information. Manual analysis of large collections of research papers is time-consuming and often leads to missed insights. Although modern AI-based systems can generate summaries, they frequently lack factual grounding and depend heavily on internet-based services. This work presents MedIngest AI, an offline biomedical knowledge engine designed to ingest PDF documents, retrieve relevant evidence, and generate citation-backed summaries. The system follows a structured pipeline that includes text extraction, chunking, embedding generation, indexing, retrieval, and summarization. A key focus is placed on maintaining evidence linkage and reliability throughout the process. The system is implemented as a modular platform with an interactive user interface that supports document management, query-based summarization, analytics visualization, and export functionality. Screens such as the document library, summarization interface, analytics dashboard, and export center demonstrate the practical usability of the system. Overall, the proposed system provides a self-contained and privacy-preserving solution for biomedical knowledge processing. It improves transparency by linking generated summaries with source evidence and enables users to work efficiently without relying on external cloud services.