Generative AI (GenAI) is poised to transform education through innovative teaching, learning, and assessment tools. This research focuses on the pivotal role of prompt engineering in enhancing the effectiveness of generative AI applications, particularly in optimizing cognitive learning outcomes. By crafting targeted prompts, educators can leverage AI to create personalized content, facilitate real-time tutoring, and provide nuanced feedback on assessments. Traditional assessment methods often suffer from inefficiencies and environmental impacts, prompting the need for smart assessment frameworks that minimize paper usage. This study proposes an AI-driven assessment model utilizing large language models (LLMs) to evaluate student responses based on semantic similarity, aligning with sustainable development goals. Our methodology integrates a generative AI stack comprising React, LangChain, and Gemini, allowing for real-time evaluation of student submissions. Assessment criteria include relevance, correctness, and depth of knowledge, enabling a comprehensive evaluation process that transcends traditional methods. The findings demonstrate that while the model achieves high relevance and faithfulness in responses, challenges remain in ensuring factual correctness. Through robust performance evaluations, this work illustrates the potential of generative AI and prompt engineering in creating a dynamic and responsive learning environment. By embracing these technologies, educators can foster improved educational outcomes, ultimately revolutionizing the assessment landscape and aligning it with sustainable practices.