Abstract: Calculus courses Math 2A/2B at UCI  enroll around 2,000 students, creating a significant grading burden for instructors and teaching assistants. In this talk, we will present an ongoing project that leverages AI to grade handwritten quizzes and exams in these courses. We will discuss the basic workflow of the AI grading system and highlight key technical challenges.  Methods to address these challenges, as well as preliminary results, related works and  future directions, will also be explored. This project aims to improve grading efficiency while maintaining accuracy and fairness, offering potential applications for other STEM courses.