This thesis investigates the relationship between ambulance crew workload and operational metrics in Emergency Medical Systems (EMS) and the applicability of machine learning models for predicting crew workload during a shift based on dispatch data. Utilizing call dispatch data from a collaborating EMS organization, the study incorporates observational task analysis, employing cumulative work time and a VACP (Visual, Auditory, Cognitive, and Psycho-motoric) scale to quantify task durations and estimate workload scores. A trace-based simulation approach is developed which integrates dispatch data and task analysis results to quantify workload. This simulation is used to study and reveal statistically significant correlations between crew workload and commonly used operational metrics, including unit hour utilization, number of calls, and scene time interval. Response time utilization, a not commonly used metric but one that can be easily estimated using dispatch data, was found to correlate better with workload than traditional metrics. Multiple machine learning models are identified from the literature review for the problem of predicting shift end workload. These models which include linear regression, decision tree regression, random forest regression, recurrent neural network (RNN), and long short-term memory (LSTM) are tuned, trained, and then evaluated in terms of performance, adaptability to input, and ease of use /applicability in EMS. A sequentially trained linear regression model is identified as the best solution for predicting ambulance crew workload at specified posts. This study lays the groundwork for future research, providing a framework to quantify medics' workload and offering data-driven insights to advance the understanding of workload dynamics for informed management decisions, potentially mitigating burnout.