About

Umang Yadav is a Senior Software Engineer at AMD, specializing in compilers for machine learning. He is primarily focused on open-sourced AMDMIGraphX project, a graph compiler for inference workloads for AMD's Data Center GPUs. Umang has a Master of Applied Science in Electrical and Computer Engineering from the University of Toronto and a Bachelor of Technology in Electronics and Communications Engineering from the National Institute of Technology Surat. With almost five years of relevant experience, Umang has worked on various aspects of machine learning acceleration on GPUs, including MIGraphX, ROCm, DirectML, and TVM exploration and comparisons. He is currently working on delivering competitive features and performance enhancements for AMDMIGraphX. Umang has previously worked as a Software Engineer at Untether AI, where he built and owned the Relay/TVM based graph transformer for the NN graph targeted towards ASIC. He also developed, enhanced, and maintained high-performance machine learning kernels for the ASIC and worked on various aspects, including quantization and performance stats. Umang also worked as an Engineer with Heterogeneous Compiler Lab at Huawei, where he worked on TVM-based compiler stack for Huawei’s ML ASIC. He developed custom loop optimizations, IR transformations, and expression simplifications for various ML ops, codegen support, and performance improvement. Umang's technical skills include software engineering, machine learning, and backend development. He has a good understanding of TVM compiler stack and has contributed to the TVM open-source project. Umang's education history includes a Master of Applied Science in Electrical and Computer Engineering from the University of Toronto, a Bachelor of Technology in Electronics and Communications Engineering from the National Institute of Technology Surat, and a High School Certificate in Physics, Chemistry, and Maths from Smt. V. D. Desai Wadiwala (Bhulka Bhavan) High School.