This dissertation proposes a learning-based, physics-aware soil moisture (SM) retrieval algorithm for NASA’s Cyclone Global Navigation Satellite System (CYGNSS) mission. The proposed methodology has been built upon the literature review, analyses, and findings from a number of published studies throughout the dissertation research. Namely, a Sig- nals of Opportunity Coherent Bistatic scattering model (SCoBi) has been first developed at MSU and then its simulator has been open-sourced. Simulated GNSS-Reflectometry (GNSS-R) analyses have been conducted by using SCoBi. Significant findings have been noted such that (1) Although the dominance of either the coherent reflections or incoher- ent scattering over land is a debate, we demonstrated that coherent reflections are stronger for flat and smooth surfaces covered by low-to-moderate vegetation canopy; (2) The influ- ence of several land geophysical parameters such as SM, vegetation water content (VWC), and surface roughness on the bistatic reflectivity was quantified, the dynamic ranges of reflectivity changes due to SM and VWC are much higher than the changes due to the surface roughness. Such findings of these analyses, combined with a comprehensive lit- erature survey, have led to the present inversion algorithm: Physics- and learning-based retrieval of soil moisture information from space-borne GNSS-R measurements that are taken by NASA’s CYGNSS mission. The study is the first work that proposes a machine learning-based, non-parametric, and non-linear regression algorithm for CYGNSS-based soil moisture estimation. The results over point-scale soil moisture observations demon- strate promising performance for applicability to large scales. Potential future work will be extension of the methodology to global scales by training the model with larger and diverse data sets.