Our Solution – On-chip Processing Using DL - We have developed low power AI accelerator core with advanced compilers and deep learning frameworks integration. 500 GOPS performance obtained in less than 1 Watt power envelope on 65 nm prototype.