Dust in the circumgalactic medium (CGM) has been detected statistically, using tens of thousands of background galaxies or quasars to measure average attenuation. However, variations in dust properties among galaxies, and the causes of those variations, are lost when stacking. To achieve a sufficient level of precision that enables spatially resolved CGM dust mapping in individual galaxies, we present a new approach applied to imaging from the Large Binocular Telescope (LBT) Low surface brightness Imaging of Galaxy Halos and their Tidal Structures (LIGHTS) survey. Specifically, we train a gradient-boosted decision tree model (XGBoost) to predict the intrinsic g − r color of background galaxies based on photometric and morphological features. By subtracting the predicted intrinsic color from the observed color, we measure the dust-induced attenuation along individual lines of sight. Leveraging this enhanced precision and deep LIGHTS imaging, we produce the first resolved 2D dust maps of the CGM in several nearby galaxies. We will discuss observed dust features and variance among galactic halos. These maps provide new avenues to study the physical conditions of the CGM, enabling investigations into dust survival, feedback processes, and the nature of the hosting dark matter potential, and to evaluate systematic uncertainties in cosmological measurements caused by an uneven foreground dust screen.