Modern computer aided design software usually has geometrical kernels developed for computer graphics purposes. Algorithms which could successfully produce high quality pictures met obstacles in modeling of solid objects. The wide-spread boundary representation produces cracks and holes in shapes performing set theoretical operations. One more task challenged widely used geometric kernels is the structural optimization or generative design. Well investigated optimization algorithms, e.g. SIMP, require to perform several conversions of the geometry representation for modeling, optimizing and manufacturing tasks. Most of them require a model to be divided into a number of space portions, which can be added or removed while the algorithm works. However, there is a geometric modeling technique which can be attractive for all mentioned routines. It is a function representation (FRep). First of all, FRep provides an ability to freely parametrize a model. Secondly, this representation guarantees the correctness of a shape for meaningful parameters. This paper contributes with the research on structural optimization of FRep models and authors’ experience of 3D printing of optimized parts. The classical compliance minimization problem is considered. Parameters introduced by a designer can be used in the optimization process to satisfy an objective function. Several techniques for lattice modeling and hole nucleation are described. Certain approaches for optimization of FRep models are discussed. Moreover, the work includes a short description of the direct manufacturing approach for FRep. The concept of direct manufacturing is realized and presented with the developed FRep system. It uses procedures for 3D modeling written in C++ language and it is integrated with third party software for producing instructions for a machine controller in an appropriate format. Finally, the obtained results of the mechanical test of a printed part are presented. This test was performed for verifying of the finite element scheme, which was used for optimization procedure.