Foundation Models (FMs) have given rise to FMware, a new generation of software now widely adopted in industrial products, such as GitHub Copilot, and across Software Engineering (SE) research. Despite this adoption, developing reliable FMware is challenging due to the stochastic nature of FMs. This tutorial presents the latest research and industry practices for engineering FMware from a pragmatic SE perspective, rather than artificial intelligence (AI), ensuring its accessibility to ASE participants. This tutorial includes novel content on recent advancements, including agentic architectures, FMware compilers, observability, and the ``FM-as-a-judge'' evaluation pattern.