The paper proposes the method and software tools for automated design and synthesis of parallel programs for field-programmable gate arrays (FPGAs) based on the algebra-algorithmic approach. The developed facilities provide the construction of parallel algorithm schemes by superposition of language constructs of Glushkov’s system of algorithmic algebra. Based on schemes, the corresponding source code in VHDL is automatically generated, which is further executed on an FPGA. The flexibility of reconfigurable FPGA architecture is very attractive for the realization of computationally complex algorithms and allows synthesizing high-efficiency solutions that differ from other architectures by substantially less energy consumption at similar performance rates. The approach to the automated design of parallel programs for FPGA is illustrated with an example of developing a genetic algorithm utilized at the training of multilayer neural networks. The results of the experiment consisting in executing the generated program code on an FPGA are given.