Can sparse neural networks make edge AI faster without increasing power consumption? An FPGA toolkit aims to prove they can.