One key parameter for estimating shale gas potential and evaluating CO2 storage capacity in shale formation is total organic carbon (TOC) content. Traditional TOC estimation methods, such as the ΔlogR technique and shallow artificial neural networks (ANN), often exhibit low correlation with laboratory measurements, limiting their applicability. This study presents a deep learning model (DLM) designed for improved prediction of TOC. The proposed model is a multilayer perceptron neural network trained using core-derived TOC values and geophysical well logs (density, resistivity, gamma ray, and caliper logs). The Pearson correlation coefficient (r) for the model predicted TOC compared with the lab-measured TOC is 0.8, demonstrating better performance than conventional techniques. The model was further trained to predict the S2 (pyrolyzable hydrocarbons) factor and enhance shale characterization. Predicted TOC and S2 values were then used to calculate the Hydrogen Index (HI) and classify kerogen type. Integrating this predictive modeling with Rock-Eval, mineralogical, and geomechanical analyses establishes a comprehensive methodology for evaluating shale gas prospects and CO2 storage potential. High TOC content (averaging 6.38%), Type III gas-prone kerogen, and favorable elastic and mechanical properties for hydraulic fracturing highlight the potential of Barren Measures shales in the Raniganj Basin, India, for shale gas exploration and CO2 sequestration.