A primary goal of live and virtual training exercises for military forces, medical practitioners, and first responders is to develop readiness in a controlled, realistic environment. Once basic skills are learned, training transfer of complex skills from exercises to a high stakes, high stress environments increases when training exercises closely mimic the stress and tempo of operations. Objectively quantifying stress in real-time would provide key data to influence and evaluate immersive training exercises. The community has investigated a number of physiological-based indicators, many of which utilize lab-based equipment that limits the ability to use such approaches in military and first responder training exercises. A stress classifier was created to allow user physiological state to be monitored and classified in real-time in an ambulatory environment, and indicated high stress classification accuracy in healthy adults and those undergoing cognitive behavioral therapy for Post-Traumatic Stress Disorder (PTSD). While the stress classifier has been assessed using both healthy and clinical populations, the ability to extend its capabilities to military and first responders was required to determine classification accuracy and ruggedness in training and operational environments. Field data collection events in collaboration with the Department of Homeland Security (DHS) First Responders Group and naval firefighter training with the Surface Warfare Officer School (SWOS) provided context relevant data to evaluate real-time stress measurement accuracy and reliability. Results from field studies in each of these two environments will be presented, along with a discussion of the findings, challenges and next steps for this technology. This paper will also summarize the unique ruggedization requirements for physiological-based algorithms and associated hardware, as well as related technologies for immersive training communities.