Laser speckle imaging (LSI) has been widely used to image blood flow in biomedical/clinical research. Commercially available perfusion monitoring systems are bulky, expensive and lack portability. Portable devices that perform LSI using custom Digital Signal Processors (DSPs), and smartphones have been proposed to perform single exposure LSI to monitor blood flow. While these can perform LSI, it is difficult to estimate flow rates using single exposure LSI. In this study, we use NVIDIA’s Xavier NX embedded vision processing platform to perform variable exposure LSI to evaluate flow rates in a 3D printed flow phantom. A low-cost Raspberry Pi HQ camera was used to acquire LSI images and the increased sensitivity of the camera allowed images to be captured with exposure times < 1ms. The higher performance of the Xavier NX platform allowed for video-rate LSI imaging (25 ± 2 FPS for spatial and 5 ± 2 FPS for temporal variance images respectively) of the phantom and using variable exposure datasets, slow (2mm/s-6mm/s), medium (6mm/s-12mm/s), and fast (>12mm/s) flow rates were distinguished. Such a platform can extend the use of portable LSI devices in peri-operative, post-operative and other point-of-care applications as this platform can be commercialized into a tablet-like form factor for convenient portable operation.