Build a custom system that utilizes the Xilinx Deep Learning Processor (DPU) IP to optimize the trained models for inference using the DNNDK targeting a Xilinx SoC.