Book excerpt: Due to the inadequate knowledge of the soil forming histories and/or human activities, the subsurface soil layers are difficult to ascertain. Subsurface uncertainty and its influence on geotechnical design have long been a challenge facing practitioners. Recently, the ASCE Geo-institute has developed the Data Interchange for Geotechnical and Geoenvironmental Specialists (DIGGS), which is a standard schema for transferring geotechnical data between multiple organizations. It paves the way of sharing and unifying datasets and forms a structural database for further data-driven modeling and analysis. The Office of Geotechnical Engineering at ODOT (OGE) is taking a national leading role in supporting the development efforts of DIGGS and hence make this project possible. In this study, site investigation data in DIGGS format and archived format are jointly processed. An innovative technique developed by the research team has been further improved for better application in real-world projects. Bayesian machine learning is integrated with Markov random field models to infer and simulate subsurface models and geospatial data with quantified uncertainty. Spatial heterogeneity and statistical characteristics are modeled in terms of statistical and spatial patterns. These patterns serve as a basis to provide a synthesized interpretation of the soil profiles with uncertainty quantified. Four (4) validation projects have been performed in this report and the results are well documented. Summary and recommendations for future work are also provided. A short introduction of the key concepts behind this technique, and pathway for converting the existing program into a ready for implementation web-based program for potential ODOT usages are provided in the appendices.