Building the best possible 3D model from observed geological data is critical before grade estimation is performed. For many decades, 3D modelling of faults has been a challenging task. Currently, explicit or implicit modelling techniques are widely used to model faults. Explicit modelling is done manually by geologists based on their knowledge and expertise. The reliability of these models depends on the geologists' abilities and creating them is a time-consuming process (Caumon et al, 2009; Wellmann and Caumon, 2018). On the other hand, implicit modelling takes a more mathematical approach to describe the fault geometry to combine the fault shape and fault frames (Grose et al, 2021; Jessell et al, 2014). Godefroy et al (2018) introduced kinematics with implicit modelling methods and directly applied them to the implicit description of the faulted surfaces. It is a challenging task to create 3D geological models comprising faults, as they represent a discontinuity in geological feature(s) that are being modelled. Geological discontinuities and their uncertainties can be easily and consistently modelled using machine learning (Sullivan et al, 2019). The underlying architecture of an implicit function has been developed and modified to provide the ability to create geological faulted 3D models in three distinct steps: data preparation; building the fault geometries; and the use of machine learning to generate complex geological faulted 3D models. This approach generates 3D models that fit the supplied data and geological knowledge of the generated fault(s), ie generates a geologically plausible faulted 3D model. The new approach is fast and repeatable compared with existing techniques.