The Arctic plays a critical role in the global climate system, but accurately predicting cloud cover and surface radiation in this region remains a significant challenge (Wei et al., 2021). These variables are crucial for precipitation prediction, temperature and humidity regulation, and severe weather forecasting. Furthermore, cloud cover directly impacts solar irradiance, making it an important factor in energy production planning for PV systems (Svennevik et al., 2021). However, many observational data sources are limited or unavailable in the Arctic, such as in-situ measurements and satellite data, which are commonly used to constrain predictions (McCusker et al., 2023). Additionally, current numerical weather prediction (NWP) systems, such as AROME-Arctic, struggle to represent the unique atmospheric conditions in the Arctic, including difficulties with temperature forecasting in cloud-free and calm condtions, solid precipitation, distinguishing freezing from non-freezing conditions, and challenges in predicting small-scale spatial variability (Wei et al., 2021) This project proposes to use machine learning to downscale coarse-resolution reanalysis data into hyperlocal predictions of cloud cover and irradiance. By integrating MODIS satellite imagery, CARRA reanalysis, and in-situ pyranometer observations, this study aims to improve the spatial resolution of predictions and contribute to Arctic climate research.