Team-based and collaborative learning have gained significant attention in educational research as they foster more profound understanding, critical thinking, and student engagement. These methods encourage active participation, facilitate the exchange of diverse skills and perspectives, and prepare students for real-world teamwork experiences, which are essential in professional settings. However, creating effective and diverse student teams is a critical challenge in face-to-face and online educational settings, particularly in large-scale environments such as Massive Open Online Courses (MOOCs) where hundreds or thousands of students can be enrolled. While several research works highlight the benefits of team-based and collaborative learning, research and technological gaps still exist in optimizing team formation and the availability and effectiveness of tools that can facilitate this process. To address these gaps, we present a novel web-based system and algorithm that facilitates the formation of heterogeneous student groups based on their skill diversity, social ties, and balanced group sizes. The system enables teachers to define activities or projects, specify required skills, collect students’ self-assessments, and visualize the student social network through a simple, easy-to-use interface. In return, students provide self-ratings on the required skills and select preferred team members. Behind the scenes, the system uses a novel hybrid algorithm integrating genetic algorithms and student social network analysis to generate diverse and balanced teams while respecting students’ social ties and preferences. The algorithm optimizes three critical objectives: maximizing skill diversity, enhancing friendship closeness and social ties within groups, and achieving balanced group sizes. Experimental results on varying-sized synthetic student networks demonstrate the system’s reliability, robustness, and stability. The algorithm consistently performs well in creating diverse groups while maintaining social ties and near-perfect size balance across clusters. The results show that the system is exceptionally well-suited for large-scale classrooms and MOOCs, enabling personalized, inclusive, and effective team creation to foster collaborative learning in diverse educational contexts.