Artificial intelligence has proven to be highly useful in various fields such as sales, finance, healthcare, and education. In the educational domain, it is applied to tasks such as predicting academic performance, student dropout, and selecting appropriate pedagogical materials. In this study, the dominant learning style among software development technology students was identified as the sequence Visual → Sensing → Active → Global, with the Visual style being the most prevalent. A multi model architecture was designed, consisting of three AI systems (two MLPs and one Random Forest) combined with six additional models, one for each dimension of the Felder and Silverman, model to address the classification problem in imbalanced datasets. Learning style prediction was performed by dimension, and the following general results were obtained: Sensing/Intuitive with F1-score = 0.85/0.17, Active/Reflective with F1-score = 0.70/0.33, Visual/Verbal with F1-score = 0.90/0.16, and Sequential/Global with F1-score = 0.49/0.38. The prediction results were promising considering that the goal is to correctly identify as many individuals from both classes as possible in each dimension, even with imbalanced data. The findings open the possibility of personalizing educational content on digital learning platforms by adapting it to the students’ learning styles, which could improve the effectiveness of teaching methodologies in technology related programs.