AI-Powered Smart Grading System Using OCR and NLP Models

Manual grading of handwritten answer sheets in educational institutions is a time-consuming and errorprone task, leading to challenges in consistency, fairness, and scalability, particularly during large-scale evaluations. To overcome these limitations, this paper presents GradeSync, an AI-powered smart grading system that automates the end-to-end evaluation of handwritten student responses. GradeSync integrates Optical Character Recognition (OCR), the Moondream Vision-Language model, and Groq AI’s natural language models into a seamless pipeline. Submitted answer sheets, captured as images, are first processed using OCR enhanced by Moondream to improve the extraction of text from diverse handwriting styles. The extracted content is then semantically analyzed by Groq AI, which compares student answers against teacher-provided keys to ensure accurate and context-aware grading. In addition to automated evaluation, GradeSync offers interactive dashboards for teachers and students, enabling access to grading summaries, personalized feedback, and in-depth academic analytics. These dashboards help educators track performance trends, identify learning gaps, and deliver targeted interventions, while students receive actionable insights to support their academic development. By streamlining the evaluation process, reducing human biases, and providing quicker feedback, GradeSync transforms traditional assessments into a more efficient, fair, and scalable system. It enables institutions to adopt a data-driven approach to academic evaluation, ultimately enhancing the overall learning experience and operational efficiency. The integration of state-of-the-art AI models in GradeSync represents a significant advancement in modernizing the grading process, making it faster, more reliable, and aligned with the growing demand for technology-driven education solutions.