The paper considers the development of an adaptive system for creating educational programs based on multilevel data analysis. The focus is on integrating labor market requirements and optimizing curricula, collecting and structuring large amounts of operational data during lessons, and adaptive testing of learning outcomes. A three-component approach is used: first, the optimization of curricula using dynamic programming algorithms, considering interdisciplinary connections and market requirements. To identify labor market requirements, surveys of employers, graduates, and teachers in the IT field were conducted. Second, as part of the real-time data processing system, the DBSCAN algorithm is used to segment students, and gradient boosting is used to personalize educational trajectories based on behavior and performance indicators. And thirdly, the improved adaptive testing system combines machine learning with Q-Learning optimization, which increases assessment accuracy by 15–20% and reduces test completion time by 50%. Experimental applications in educational institutions demonstrate significant improvements in student performance and curriculum relevance to labor market needs.