With the increased use of smart systems in the field of education, came the opportunity to optimize the process of assessment, make it more automated. The traditional way of looking at a subjective response can be time consuming, laborious and prone to inter-rater variance. The present study suggests that a bot model based on the usage of Large Language Models (LLM) can be designed to develop an intelligent feedback, and they can be used for automating the process of assessment of subjective answers. There is the inclusion of the use of advanced natural language processing (NLP) to pinpoint the context, semantic meaning and depth of knowledge in student responses. Through the fine-tuning of the LLM, the system has the ability to offer constructive, rubric aligned, and personalized feedback as well as fairness and scalability in the assessment as well. Besides, the explainability mechanisms as well as the correspondence to the standards of the pedagogical procedure are also offered in order to develop an environment that builds transparency and trust in automated grading. The given system can make a tremendous impact in terms of reducing the evaluator work, enhancing the learning outcomes due to the timely feedback, and supporting the bigger vision of AI-driven personalized education.