Here we present the idea of design of architectured lattice materials guided by analysis and optimization of the lattice structure at mesoscale. It is based on the mesoscale unit cell (greater than the unit cell of the lattice itself) approach whereby a lattice pattern is formed by different elements. This approach achieves two main goals. Additional variable design parameters allow one to significantly expand the range of possible properties. Moreover, it enables flexible control of the number of variable parameters, which, together with the use of machine learning methods, makes it possible to explore the multidimensional structure-property space and to find in it the mesoscale lattice pattern that provides a desired combination of properties.