Analyzing complex systems is a challenging process which requires not only teams of domain experts but often also a multidisciplinary team of data scientists, mathematicians, statisticians, and software engineers in order to support the life cycle of model development, model-based inference, information extraction, machine learning, and knowledge synthesis. The models that typically result from this process today are bespoke, lack generalizability, are not performable, lack reusability, and make the task of synthesizing actionable knowledge and policies from their raw outputs difficult. These problems are exacerbated by the complexity and domain intersection involved in multi-scale models, such as those used for research and diagnosis of human health problems.

We present our work on AMIDOL: the Agile Metamodel Inference using Domain-specific Ontological Languages, a project that aims to reduce the overhead associated with the development, deployment, maintenance, and reuse of multi-scale models of complex systems. Our technique utilizes a common intermediate representation which is designed to support a number of scientific, physical, social, and hybrid domains by allowing domain experts to define their models in a novel way: using domain specific ontological languages (VDSOLs). The intermediate representation provides formal, executable, meaning to the semi-formal diagrams domain experts normally create, and allows the inference engine to build prognostic queries on associated reward models. AMIDOL binds results from the inference engine to the original ontologies providing more explainability when compared to conventional methods.