This work presents a Machine Learning–based automated answer sheet evaluation system designed to improve the efficiency, consistency, and fairness of academic assessment. The system integrates Optical Character Recognition (OCR) to convert handwritten responses into digital text, Natural Language Processing (NLP) techniques for text preprocessing, and semantic similarity algorithms to evaluate conceptual correctness. The architecture includes modules for image preprocessing, handwriting recognition, NLP normalization, semantic analysis, and a machine learning scoring engine. Transformer-based embeddings such as BERT and similarity metrics like cosine similarity and Jaccard index are used to compare student answers with model solutions. The proposed approach reduces manual grading workload, minimizes human bias, and provides faster, structured feedback. It is scalable for use in schools, colleges, and large-scale examinations and supports multiple answer formats and languages. This research contributes to intelligent educational assessment systems by combining OCR, NLP, and machine learning for automated evaluation.