Capillary pressure (Pc) measurements, together with conventional core analysis, are typically used for reservoir characterization and saturation height modeling (SHM). Workflows involving Pc data are time consuming and interpretively biased. We aim to enhance these workflows’ efficiency and reduce the interpretation bias by automating the Pc-based pore network characterization and applying machine learning (ML) to capillary pressure modeling. We have also built advanced analytics dashboards to allow for QC and interactive adjustments by the user. The solution defines the pore system modality based on the automated peak(s) detection of the pore throat radius (PTR) distribution and computes an absolute porosity distribution and its porosity partition. Next, it builds the SHF per rock type at special core analysis (SCAL) level using a novel approach based on a symbolic regressor logic. This automatically defines the best correlation of the extracted fitting coefficients to core porosity and permeability while honoring the physical reservoir constraints. Furthermore, the system applies analytical equations with fitting coefficients replaced by defined correlations to capillary pressure data and automatically identifies the equation and correlations that provide lower relative error to the input capillary pressure curves. We developed and tested the proposed algorithms on a real data from a large carbonate oil field in the Middle East. The input dataset is rich, consisting of around 600 good to excellent quality mercury injection capillary pressure (MICP) samples. Parent-plug porosity and permeability is also available for all samples. Initial experiments demonstrated that the proposed workflow significantly improves the Pc-based pore network characterization workflow. Initial QA/QC algorithms based on advanced data science outliers’ detection logic allows us to automatically identify poor quality data and outlier samples. These outlier samples are filtered out eliminating data-quality-artifacts in the interpretation. The automated pore network characterization solution allows a clear understanding of the porosity partition and modality in few steps. This characterization will be the foundation of the petrophysical grouping at MICP level. We present a thorough comparison between results from petrophysical rock typing using traditional approaches compared to results obtained using the newly proposed pore network characterization methodology. Although the focus of this study is to improve the petrophysical grouping component of the rock-typing, further integration with geology and diagenetic overprints is required to improve the over-all workflow. Moreover, we automatically define the best Pc analytical model and compute the final saturation height functions using ML algorithms. This novel workflow allows more complex mathematical modelling compared to the currently available traditional approaches in commercial petrophysical packages. The overall results provide an enhanced mathematical solution constrained to the petrophysical model for the Core Build Model (CBM).