Artificial Intelligence (AI) is transforming engineering education by enabling adaptive learning, real-time feedback, and scalable instructional support. This paper explores the role of AI as a virtual teaching assistant through the integration of Intelligent Tutoring Systems (ITS) and adaptive feedback mechanisms. We propose a modular, multi-layered architecture that incorporates learner profiling, domain-specific ontologies, pedagogical strategies, and reinforcement learning to deliver personalized and context-aware instruction. Case studies from platforms such as Carnegie Learning, Squirrel AI, and Codio demonstrate significant improvements in student engagement, accuracy, retention, and time efficiency. These systems address instructional gaps, particularly in large or hybrid classrooms, by dynamically adjusting content and feedback based on real-time learner data. Beyond the technical framework, we address critical implementation challenges including faculty readiness, the need for explainable AI, fairness in algorithmic feedback, and student data privacy. We advocate for a human-in-the-loop approach, positioning AI as an instructional partner rather than a replacement, to support ethical and pedagogically aligned learning environments. The paper concludes with strategic recommendations for future curricula, highlighting the importance of interdisciplinary faculty training, robust AI governance, and longitudinal assessments. Our findings underscore the potential of AI-powered tutoring systems to enhance equity, scalability, and personalization in engineering education.