Views of theory structure in philosophy of science (semantic and syntactic) have little to say about how theories are actually constructed; instead, the task of the philosopher is typically understood as reconstruction in order to highlight the theory's essential features. If one takes seriously these views about theory structure then it might seem that we should also characterize the practice of building theories in accordance with the guidelines they set out. Examples from some of our most successful theories reveal nothing like the practices that conform to present accounts of theory structure. Instead there are different approaches that partly depend on the phenomena we want to account for and the kind of theory we desire. At least two strategies can be identified in high energy physics, (1) top down using symmetry principles and (2) bottom up strategy beginning with different types of models and gradually embedding these in a broad theoretical framework. Finally, in cases where methods and techniques cross disciplines, as in the case of population biology and statistical physics, we see that theory construction was largely based analogical considerations like the use of mathematical methods for treating systems of molecules in order to incorporate populations of genes into the theory of natural selection. Using these various examples I argue that building theories doesn't involve blueprints for what a theory should look like, rather the architecture is developed in a piecemeal way using different strategies that fit the context and phenomena in question.