Amin Ansari

Experience

1. Senior Staff Machine Learning Engineer

   高通

   Jan 2013 — Present

   ADAS/Autonomy Systems R&D Team Working / leading / publishing on a variety of problems related to the application of machine learning and deep learning to computer vision for autonomous driving. I particularity have worked on object detection using camera (2D, 3D, and BEV) and radar (RAD cube and point-cloud) signals, worked on the early-fusion of radar and camera sensors, semantic segmentation, monocular dense depth estimation (self-supervised and partially-supervised), multi-view depth estimation, automatic cross-sensor calibration (between camera and LiDAR), domain adaptation to improve long-tail problem using a hybrid of simulation and real-world data, and convolutional and recurrent neural networks exploiting temporal information to improve the quality of the underlying perception module. • System-on-Chip Architecture Research Team Worked on accelerators for deep convolutional neural networks. I have studied the evolution of networks, worked on the compression of the weights and activations, and designed an architecture to exploit sparsity. I also worked on camera image processing pipeline as well as on designing a multi-purpose, lossless, on-chip compression engine to reduce the memory footprint and improve the system-on-chip bandwidth. In addition, I designed and implemented a bandwidth reduction technique for GPU to reduce power consumption and improve overall throughput.