As technology progresses, online platforms are becoming increasingly important in various sectors, including testing. While traditional pen-and-paper tests have long been the standard, recent events like the global pandemic have accelerated the shift to online evaluation methods. Digital multiple-choice question (MCQ) formats and handwritten responses submitted via smartphones are now prevalent. However, effectively grading theoretical subjects remains a significant challenge. This research introduces an innovative method using Natural Language Processing (NLP) for automated grading of handwritten answers. We employ Optical Character Recognition (OCR) algorithms to convert handwritten text from test papers into a digital format. The grading system compares vector embeddings of student responses with those of instructors. Higher grades are assigned when the vectors are closely aligned, indicating a strong similarity. By integrating NLP and OCR technologies, our proposed method aims to streamline the grading of handwritten answers in theoretical subjects. This approach addresses the evolving needs of online testing systems while maintaining accuracy and efficiency. The results indicate that this method can uphold grading quality in a digital and remote learning context.