Techniques to perform code separation with semantic guarantees are described. When porting a software application to a different target architecture and potentially to a different set of platforms and programming languages, application functionality will change based at least to accommodate architectural differences. However, business logic should be guaranteed to be semantically preserved, efficiency optimized to the platform benefits and limitations. Code separation identifies candidate components, associates the candidate components with a component classification or type, and automates the determining of which of the candidate components are to be associated with various portions of the target architecture. Automated code separation may include various inference algorithms to determine which components are to be associated with a component type. Also described are various use cases for code separation.