Digital Peer-Reviewer with LLM-integration

Abstract

This project develops an AI-powered web application to automate academic peer review processes and assess research novelty. Built on the MERN stack, the system integrates DeepSeek’s language model to generate structured reviews and utilizes PubMed’s database to identify similar research through keyword extraction and semantic analysis. Users upload PDF documents, which are processed to extract metadata and text content, with cached results reducing redundant computations. Security is maintained via Azure’s isolated virtual machines and encrypted communications. The application successfully retrieves relevant prior research in 92% of test cases and generates reviews aligned with human feedback in critical areas. The interface organizes results into digestible sections for methodology evaluation, originality insights, and improvement suggestions. While limited to English-language text and PubMed-based comparisons, the system demonstrates potential to streamline peer review workflows through automated analysis. Future expansions could address multilingual support and broader literature databases.