AAI_2025_Capstone_Chronicles_Combined

4

especially in settings with high student-teacher ratios, where providing timely feedback is challenging (Binfoh Abuaa et al., 2022). It can be used to provide multiple explanations for math concepts, tailoring learning to students with different skill levels and backgrounds. Through guided explanations and personalized support, AI could help students develop a better understanding of fundamental mathematical concepts (Mohamed et al., 2022), cope with self-efficacy and anxiety in math (Inoferio et al., 2024), and engage more in math (Richard et al., 2022). Beyond supporting learning, virtual teaching assistants can collect student data to generate performance profiles that reveal individual and class-wide trends, enabling teachers to pinpoint student needs and adjust lesson priorities (Audras et al., 2022). When this data is made readily available, teachers can use it to make prompt instructional decisions. This research aims to design and implement an AI-driven education platform for tutoring students in middle school mathematics. A key goal of the platform is to enhance the learning experience for students by providing guidance that is customized to both the course curriculum and the learning style of each student. This guidance should develop the analytical reasoning skills of students and help create a foundational understanding of mathematics when used as a supplement to in-class instruction. Through the use of scaffolding, incorporation of real-world examples, and providing of emotional support, we hope to encourage students’ self-assurance and intellectual curiosity when learning mathematics. The developed platform is an intelligent tutoring system that ingests information about the course curriculum and creates sets of guided questions for each lesson, letting students converse with a chatbot tuned for educational contexts about key concepts, practical examples, and related topics for that lesson. To address the potential for Large Language Models (LLMs) to generate inaccurate or ungrounded responses, we integrate retrieval-augmented generation (RAG) to provide additional grounding for the tutor, helping ensure that responses are aligned with the course materials. A teacher dashboard gives instructors the ability to explore insights from student chats at both the lesson level and the individual student level. These insights help instructors quickly identify valuable classroom information such as common question topics,

228