The term “learning style” refers to an individual’s preferred approach to acquire knowledge. Most learning systems of today fail to identify the learning differences among students; and group them together offering the same content to all learners. This leads to disorientation and cognitive overload problems among learners. The solution is to provide personalized and adaptive learning content as per a student’s learning style. Learning theories such as VARK suggest that a student’s learning style could be either Visual, Auditory, Read/Write or Kinaesthetic. On the contrary, we suggest here that any student cannot just have a single learning style but would have a combination of multiple learning styles; though the proportion would vary amongst different students. Given this, it seems an enormous problem to create and maintain personalized learning content unique to each student. We address this concern here and provide an approach to cluster students into an optimal number of classrooms based on their learning style preferences. Data is obtained by circulating a VARK model-based questionnaire among 200 engineering students. Different clustering techniques i.e. K-Means, Hierarchical, Fuzzy C-Means and DBSCAN are applied on the data and results are compared using various metrics like Silhouette Score, Davies-Bouldin Index (DBI) and Calinski-Harabasz Index (CHI). The best results are obtained with K-Means clustering. Eventually, three clusters are obtained for a group of 194 students, each with a unique combinations of learning styles. This clearly demonstrates the importance of grouping students into different classrooms as per their learning style combinations; so that personalized learning content can be delivered to them.