Optimizing talent supply chains: deep learning models for resume-based retention forecasting

Abstract In today’s dynamic labor market, organizations face growing challenges in attracting and retaining top talent. High employee turnover imposes substantial costs and disrupts strategic workforce planning. The aim of this paper is to investigate where predictive modeling, specifically deep learning, can be applied to improve recruitment strategies by forecasting candidate retention time, based only on resume data. Conducted in collaboration with Talendary, an AI-driven recruitment platform, the study presents a neural network regression model trained on over 90,000 anonymized resumes. Utilizing principles from Supply Chain Management, the model incorporates features such as career progression rate, job stability, and mobility patterns. The results demonstrate strong predictive performance, achieving an $$R^2$$ score of 0.9877 on the test set, highlighting that resumes carry meaningful signals about future job tenure. These findings suggest that data-driven recruitment, supported by machine learning, can enable more informed hiring decisions, reduce turnover-related costs, and contribute to long-term workforce sustainability. Ethical considerations, including fairness, transparency, and privacy, are also addressed to support responsible AI use in hiring processes.