This study aims to precisely quantify soil carbon stocks in nine farms in the Red River Valley using tools like precision agriculture, GIS, remote sensing, and deep learning. Over 4,000 soil samples were taken and analyzed for total carbon, inorganic carbon, organic carbon, and bulk density. Remote sensing data and yield data were also collected. Different machine learning models were tested to predict carbon levels from the data, with gradient boosting and histgradient boosting models achieving R2 values around 0.8. Traditional methods achieved lower accuracy. The research is 40% complete and aims to finish within three years to allow for better carbon sequestration contracting and investment targeting.