Over the last decade artificial neural networks (NN) have been proposed as alternatives to more conventional elasto-plastic based constitutive models. A NN material model can be used in a similar fashion to conventional models in the solution of boundary value problems. The authors have developed Self Learning in Engineering Simulation analysis framework which extracts relevant soil behavior using boundary measurements of load and displacement, facilitated by use of NN constitutive model. This framework shifts focus from the specific soil model to the stress-strain database used in material model training and development.