Previous studies have confirmed the pivotal roles of student engagement in predicting learning achievements; however, only a few emphasized the patterns of different engagement dimensions within individuals (i.e., co-occurrences of behavioral, cognitive, and emotional engagement). Thus, this study identifies engagement profiles and explores their potential antecedents (i.e., achievement goal orientations, academic self-efficacy, and perceived task difficulty) and consequences (i.e., task performance). One hundred and eleven (N = 111) medical students solved a diagnostic task in BioWorld, a technology-rich learning environment that simulated virtual patients to improve students' diagnostic skills. Behavioral engagement was extracted from computer trace data, whereas both cognitive and emotional engagement were measured by questionnaires. Using a latent profile analysis (LPA), this study revealed four engagement profiles: Cognitively Disengaged, Behaviorally Engaged, Moderately Engaged, and Cognitively and Emotionally Engaged. Subsequent multinomial logistic regressions showed that mastery-approach goals and academic self-efficacy significantly increased the likelihood of being memberships in the Cognitively and Emotionally Engaged profile. Moreover, higher perceptions of task difficulty led students to be more cognitively and emotionally engaged. Additionally, students in the Cognitively and Emotionally Engaged profile tended to achieve a higher diagnostic accuracy (i.e., whether students achieved a correct diagnosis), whereas Behaviorally Engaged students were more likely to obtain higher scores in diagnostic efficacy (i.e., the overlaps between students' and medical experts' diagnostic processes). These findings support the multidimensional engagement perspective, and the implications of these profiles for students' motivation and instructional designs are discussed.