We demonstrate a novel method to model finite metaclusters that can steer the energy of an incident wave preferentially toward a given direction. This is accomplished by solving an inverse multiple scattering problem for selecting a desired energy distribution of scattered waves. One can steer the incident energy toward a desired direction using a 2-D configuration of metacluster with finite number of fluid cylinders embedded in a homogeneous fluid medium. For a faster implementation of the method, we approximate the problem to small cylindrical particle limit which corresponds to low frequency scattering. The required mechanical properties of fluid scatterers are defined by T-matrix components obtained by solving linear system of equations. A major challenge in implementing our computational model and applying to design of metaclusters devices is ensuring that the scatterers remain manufacturable using available conventional materials with positive values of mechanical properties. Metaclusters are designed by minimizing the relative error between a given and computed scattering patterns and using advanced optimization algorithms and deep learning.