Abstract
The article investigates the theoretical and practical foundations of comprehensively improving computer science teachers' professional training through the integration of STEM education and machine learning (ML). A three-level competency development model is proposed, grounded in the TPACK framework and constructivism theory. A system of five practical task blocks directly mapped to school computer science curricula is described and experimentally validated. The experimental study, conducted at L.N. Gumilyov Eurasian National University over one semester with 80 pre-service teachers, demonstrates statistically significant gains across all measured competency indicators (p < 0.05). The study provides specific, implementation-ready recommendations for integrating STEM-ML modules into teacher education curricula and professional development programs in the Central Asian context.