With the increasing risk of vehicle theft, break-ins, and vandalism, the need for effective sentinel monitoring has become paramount. However, maintaining continuous surveillance around the parked vehicle without rapidly depleting the battery is critical. While high-power sentinel monitoring is currently used in battery electric vehicles (BEVs), a significantly lower-power solution is needed to enable its adoption in hybrid and internal combustion engine (ICE) vehicles. This presentation introduces three low-power architectural approaches for sentinel monitoring, including a solution that integrates a neural processing unit (NPU) at the edge. By enabling ultra-low-power intelligent sensing at the sensor edge and leveraging specialized capabilities over automotive high-speed video links, this architecture cuts sentinel monitoring power consumption by over