With increasing challenges in climate change and extremes, monitoring the Earth’s surface with enhanced temporal availability on a global scale is of vital importance. Utilizing reflected L-band navigation signals as signal-of-opportunity, Global Navigation Satellite System Reflectometry (GNSS-R) has emerged as a promising remote sensing technique for retrieving surface parameters. Several spaceborne satellite missions, e.g., NASA’s CYGNSS and ESA’s PRETTY, offer an unprecedented sampling rate on the order of millions of measurements per day, demonstrating strong capabilities for versatile applications.The rapidly growing amount of GNSS-R data has sparked a growing interest in data-driven approaches within the GNSS-R community. Owing to the powerful ability of deep learning methods to learn underlying mappings between different geophysical parameters, recent advances have explored their effectiveness and enhancement in monitoring ocean wind speed, sea ice, surface soil moisture, inland water bodies, and vegetation.Within the scope of our AI for GNSS-R project, we aim to exploit the synergy between massive amounts of GNSS-R data and state-of-the-art deep learning algorithms. Our studies have demonstrated enhanced wind speed estimation compared to conventional and other deep learning methods, especially for high wind regimes and during precipitation events. We have also explored the potential for daily measurements of vegetation states that provide valuable insights into the Earth’s hydrological and carbon cycles, filling measurement gaps in current products. Finally, we investigated estimation uncertainty and feature importance of various geophysical parameters to help understand the reliability of measurements and the decision-making process of data-driven methods.