Perception devices, such as AR/VR wearable devices and autonomous vehicles, undergo deformations due to temperature changes, mechanical stress and aging, among other factors. This degrades the user experience as it reduces the accuracy of mapping and tracking, which depends on accurate calibration. In wearable AR devices, miscalibrations are especially jarring since the user sees virtual content overlaid on the real world. This means that maintaining accurate calibration is essential for these devices. To do so, devices must be re-calibrated frequently. Traditionally this has been done by repeating external calibration, which is an involved procedure and often requires special equipment. In this talk, we present key aspects of our work in developing and deploying advanced auto calibration algorithms that are fast, accurate and reliable—and that run on embedded hardware while avoiding adding any tasks for the user.