Abstract: Research has demonstrated that the connection between engagement and success is very critical in a learning environment. Engaging students in the learning process enhances their attention and heightens their focus, stimulates their higher critical thinking skills, and bolster meaningful learning. At the same time, emotions play an essential role in the learning activity. If a learner is not in a state of learning due to fear, frustration, anxiety, and or depression, it can lead to boredom and non-clarity. The intelligent tutoring system (ITS) is designed to simulate a human tutor in a student-centered adaptive teaching and learning environment. It incorporates artificial intelligence by invoking deep machine learning and neural networks to deliver a customized lesson to the learner. The ITS registers student's prior knowledge on the subject, learning habits, and styles in its log files, and develops strategies for teaching and engaging students while delivering the lesson. It also keeps track of the student's gaze patterns on the computer screen. In case of student's boredom, disengagement, or zoning out, the Digital Tutor (DT), a virtual pedagogical agent of the ITS, engages the learner using two-way dialogs. This paper discusses the Digital Tutor's design challenges due to student disengagement and strategies to re-engage the learners using gaze patterns, emotion recognition, voice recognition, and two-way dialogs using voice cloning.