Development of a Socio-Technical Deep Neural Network Framework for Personalized Career Guidance and Entrepreneurship Training

The increasing complexity of modern labor markets and entrepreneurial ecosystems demands service systems that adapt to users' multidimensional profiles. Drawing on a systems approach and socio-technical theory, the author proposes a unified architecture for personalized career guidance and entrepreneurship training. A multilayer network encodes heterogeneous data—career behaviors, ability assessments, and interest embeddings—for unified feature representation. A multitask optimizer combining cross-entropy and weighted regression losses enables end-to-end training for career-direction recommendation and ability-growth-path prediction. The system comprises a real-time feature collection module, a model inference engine with feedback loops, and a closed-loop interface for multi-round dynamic recommendations. Experimental results on entrepreneurial project managers demonstrated 94.0% accuracy in career-direction recommendations, a mean squared error of 3.09 for ability-path prediction, and 89.5% top-10 course coverage.