This study develops a discrete-event simulation model using open-source software to analyze import container flows through the Port of New York/New Jersey. The simulation integrates parameters from extensive data analysis of vessel arrivals, container dwell times, and intermodal transfers. Data is sourced from multiple public sources including vessel GPS data and import records released by Customs and Border Patrol. The model is calibrated with real-world import data to ensure accuracy. A fine-tuned BERT model is used to predict Harmonized System (HS) codes using unstructured shipping manifest text, achieving over 90% classification accuracy at the two-digit level. This classification enables commodity-specific analysis of dwell time. Dwell time analysis results show that for dry containers, Gaussian KDE reduces mean absolute error by 39.5% relative to the normal distribution. For refrigerated containers, the Fourier Series model yields a 24.8% reduction. Scenario testing allows us to quantify the effects of changes in key variables such as container arrival volumes, resource allocation, and truck-rail modal split – performance metrics like throughput, truck congestion, and container dwell times are captured. Scenario testing reveals that increasing the outbound rail share from 15% to 25% reduces truck congestion by 11.5% and reduces median dwell times by 1.52% for dry containers and 2.55% for reefers. Extending gate hours by two hours per day leads to a 4.42% drop-in median dwell time for dry containers and 6.28% for reefers. The model provides a scalable, data-driven framework for evaluating operational and policy interventions. Further research can expand on short-haul rail utilization, extended gate hours, or inland infrastructure investments.