This study uses machine learning to automate aspect-based sentiment analysis (ABSA) of Vietnamese employee reviews. The data comes from two Vietnamese company review platforms. After preprocessing and manual labeling of 1268 employee reviews across 10 aspects (based on crawled data), two BERT-based architectures were fine-tuned and evaluated. They achieved a promising accuracy of 0.79, confirming the method’s feasibility on the combined dataset. The research shows that model performance is affected by data source characteristics. It also highlights the dominant positive (relationships, workload) and negative (organizational structure, advancement, leadership) aspects. The model provides the foundation for developing a data-driven analytical tool that helps businesses understand employees, supports HR management and builds their employer brand.