AUTOGRAD

Abstract: The increasing number of student assessments in educational institutions has made manual evaluation a time- consuming and resource-intensive process. AutoGrad is an Artificial Intelligence-based Automatic Grading System designed to automate the evaluation of handwritten and digital answer sheets while maintaining accuracy, consistency, and fairness. The system integrates advanced technologies such as Optical Character Recognition (OCR), Natural Language Processing (NLP), Machine Learning, and Large Language Models (LLMs) to assess student responses efficiently.
AutoGrad begins by accepting scanned answer sheets through a secure web interface. The uploaded documents undergo preprocessing techniques, including image enhancement, noise reduction, and segmentation, to improve recognition quality. A Gemini-based OCR engine extracts textual content from handwritten answers, while NLP techniques and semantic analysis evaluate the responses against predefined model answers and grading rubrics. The system further employs vector similarity matching and contextual reasoning using a fine-tuned language model to assess answer quality, completeness, and conceptual correctness.
In addition to automated scoring, AutoGrad generates detailed feedback and performance reports for students and educators. The platform supports objective, subjective, and numerical answer evaluation, making it suitable for a wide range of academic assessments. By reducing faculty workload, minimizing human bias, and delivering rapid results, AutoGrad enhances the efficiency and transparency of the evaluation process.
The proposed system demonstrates how modern AI technologies can transform traditional assessment methods and contribute to the development of scalable, intelligent, and student-centric educational environments.

Keywords: Automatic Grading System, Optical Character Recognition (OCR), Natural Language Processing (NLP), Large Language Models (LLMs).