Modular organization is a special feature shared by many biological and social networks alike. It is a hallmark for systems exhibiting multitasking, in which individual tasks are performed by separated and yet coordinated functional groups. Understanding how networks of segregated modules develop to support coordinated multitasking functionalities is the main topic of the current study. Using simulations of biologically inspired neuronal networks during development, we study the formation of functional groups (cliques) and inter-neuronal synchronization. The results indicate that synchronization cliques first develop locally according to the explicit network topological organization. Later on, at intermediate connectivity levels, when networks have both local segregation and long-range integration, new synchronization cliques with distinctive properties are formed. In particular, by defining a new measure of synchronization centrality, we identify at these developmental stages dominant neurons whose functional centrality largely exceeds the topological one. These are generated mainly in a few dominant clusters that become the centers of the newly formed synchronization cliques. We show that by the local synchronization properties at the very early developmental stages, it is possible to predict with high accuracy which clusters will become dominant in later stages of network development.