At any given moment, a patient with posttraumatic stress disorder (PTSD) can be reminded of prior trauma and immediately overcome by debilitating stress. A key challenge in mitigating traumatic stress is that trauma recall generally takes place outside the clinic - away from clinicians who would otherwise be able to detect and intervene to help curb the response. This dissertation describes biosignal processing and dynamical modeling efforts to enable closed-loop technologies for traumatic stress. These systems will be able to measure changes in stress using non-invasive physiological sensing and curb pathological stress using non-invasive vagus nerve stimulation. Toward this goal, we describe our efforts to quantify changes in stress - on the order of seconds - using physiological markers (i.e., ``physiomarkers"). We then characterize these markers' responses to transcutaneous cervical vagus nerve stimulation (tcVNS) and trauma recall to assess the viability of tcVNS as a stress-reducing intervention. Through the research described, we make the following key contributions. First, we detail our design and validation of robust biosignal processing pipelines for cardiovascular and respiratory signals and show that these pipelines can extract physiomarkers pertinent to trauma recall and tcVNS effects. We then demonstrate that by quantifying breathing irregularity using physiomarkers of respiratory variability, we can capture variations in neurophysiological and psychological symptoms of trauma recall. Then, we investigate the effects of tcVNS by elucidating the specific time courses of two cardiovascular physiomarker responses to tcVNS. Importantly, we demonstrate that tcVNS effects occur within 10-15 s. Finally, we build on these single-marker modeling efforts by characterizing the way nine cardiovascular and respiratory physiomarkers covary in response to trauma recall and tcVNS. We demonstrate that although the dynamics in response to trauma recall are faster and larger in magnitude, the dynamics in response to tcVNS oppose these effects. Notably, the learned principal axis mirrors what is expected physiologically during changes in stress, and this is captured entirely unsupervised.