This thesis explores optimizing the arrangement of buildings on a site according to a set of spatial constraints. Design objectives include solar insolation and building proximity. The difficulty of this task increases considerably with the number of buildings on a site or the complexity of site and building geometries. The solution space associated with such a problem is typically infinitely large and therefore its exploration could be supported by design optimization methods. For that purpose, an optimization-based site planning system was designed and implemented. The system was implemented in an existing geometric design software that provides optimization functionality based on genetic and simulated annealing algorithms. The system was further improved using fine-tuning procedures, involving experiments with the different algorithms and with different implementation decisions. The efficiency of the system and the quality of the results produced is evaluated in two ways: By producing a series of optimization examples according to different system settings for an arbitrary set of buildings and site and by simulating an existent design problem for two case studies. After the simulation is completed, the values of the design criteria of each case are calculated for the actual buildings and compared to the ones produced by the system.