With the expansion of China’s postgraduate education and the intensification of structural contradictions in regional employment, accurately predicting graduates’ career destinations holds significant practical value for optimizing the synergy between talent cultivation and regional strategic demands. This research focuses on master’s degree graduates in life sciences from the University of Electronic Science and Technology of China (UESTC), integrating multidimensional feature data, including disciplinary background, place of origin, and research team affiliation, to construct a machine learning-based predictive model for geospatial career choices. To address data sample imbalance, logistic regression and random forest algorithms were used to optimize the model. Model performance was comprehensively evaluated using accuracy, precision, F1 score, and area under the curve (AUC) metrics. The findings suggest that disciplinary specialization and the tier of hometown cities are influencing factors in the decision-making process of life science postgraduates regarding domestic or overseas further education, whereas geospatial preferences for direct employment are jointly shaped by team ecosystems and regional industrial advantages. Innovatively, this research establishes a dynamic “cultivation-employment” prediction system, achieving a paradigm shift from “postemployment feedback” to “precultivation intervention.” It provides data-driven decision support for universities to proactively adjust disciplinary structures and for government agencies to formulate regional talent policies.