This document proposes using machine learning and sentiment analysis to predict employee service time. It introduces the problem of staff attrition rates affecting company performance. It then describes analyzing a dataset of over 36,000 examples with 10 variables related to federal employee records to build machine learning models for predicting average length of service. The models were evaluated on accuracy and AUC, with AdaBoosting performing best at 76% accuracy. Sentiment analysis of online comments found 78% supported extending work visas for STEM students. In future work, the authors aim to expand topics covered and optimize parameters to maximize company benefit.