This chapter explores the application of artificial intelligence (AI) for automatic scoring in STEM education, addressing the challenges of data diversity, contextualization, fairness, and scalability. A comprehensive framework is proposed, covering essential components such as data collection, preprocessing, algorithm selection, and model evaluation. The chapter presents three use cases: (1) pretraining the Transformer language model; (2) fine-tuning ChatGPT for fair automatic scoring; and (3) efficient and scalable scoring, illustrating the potential of AI to enhance assessment accuracy and reduce grading burdens. By leveraging advanced techniques like pretraining, fine-tuning, and knowledge distillation, this chapter provides a roadmap for researchers and educators to develop robust, equitable, and accessible scoring systems. Future directions underscore the need for diverse datasets, ethical frameworks, and explainable AI to ensure the responsible deployment of AI in education.