Project Overview
PyTutor is an AI-powered tutoring platform develoepd through a collaborative effort between Massachusetts Institute of Technology (MIT), Georgia State University (GSU), and Quinsigamond Community College (QCC). The platform is designed to deliver real-time, pedagogically sound, and personalized tutoring support for students across a range of courses and topics.
The platform is built as an interactive learning environment, supporting components such as an executable code editor and a collaborative whiteboard. At its core is a context-aware AI tutor. Powered by a large language model, it leverages three sources for context: course content (e.g. lecture materials and assignments), student’s ongoing work, and past interactions with the tutor.
To ensure effective tutoring, researchers of the PyTutor team has explored various approaches, such as advanced prompt engineering and reinforcement learning, to generate responses that aligns with pedagogical principles. PyTutor also tracks student progress throughout the course to personalize according to the student’s learning path and preferences.
Designed for scalability and adaptability across a wide variety of subjects, PyTutor allows instructors to tailor the student’s workspace setup as well as the tutor’s behavior using course-specific materials and configurable settings. It has been successfully deployed in introductory Python programming courses across all three partner schools, and is now expanding to support high school calculus, with enhanced multimodal tutoring capabilities.