Ashutosh Parkhi
Principal Software Engineer at AMD
Cambridge, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolAshutosh Parkhi is a Principal Software Engineer based in Cambridge with a strong track record of building and leading teams to deliver ML compiler and accelerator integration for industry-leading IPs. Over a 15+ year career spanning Intel, Imagination, Arm and now AMD, he has combined hands-on compiler engineering with people leadership—leading PyTorch/TFLite compilation efforts and hiring at Arm’s NPU teams. He is an active contributor and Apache TVM committer, notably implementing and optimizing Arm CMSIS-NN support (codegen, graph partitioning and efficient Conv2D/Depthwise/Softmax/FC implementations). Ashutosh excels at mapping heterogeneous neural workloads across CPU, GPU and NPUs, having designed SDKs and flows that unify TensorFlow/Keras/TFLite to lower-level IRs like TOSA and IREE. He brings both deep systems-level simulation experience and practical delivery of production SDKs, and often bridges research proofs-of-concept with shipping software. His background (MSE University of Michigan, IIT-BHU) underpins a pragmatic approach to complex compiler and accelerator co-design.
4 years of coding experience
13 years of employment as a software developer
Master’s Degree, Computer Engineering, Master’s Degree, Computer Engineering at University of MichiganBachelor’s Degree, Computer Engineering, Bachelor’s Degree, Computer Engineering at Indian Institute of Technology (Banaras Hindu University), Varanasi
Github Skills (10)
cmsis10
compiler10
machine-learning10
compiler-compiler10
deep-learning10
n10
performance-monitor9
performance-analysis9
tensor8
gpu6
Programming languages (5)
Github contributions (5)
apache/tvm
Aug 2021 - Jan 2023
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
ML Engineer
ML EngineerContributions:278 reviews, 51 commits, 84 PRs in 1 year 4 months
Contributions summary:Ashutosh's primary contributions revolve around integrating and optimizing machine learning models within the `apache/tvm` repository, specifically focusing on the Arm(R) CMSIS-NN framework. Their work includes implementing and refining support for CMSIS-NN, including code generation, graph partitioning, and optimizing operations like Conv2D, Depthwise Convolution, Softmax, Pooling, and Fully Connected layers. The user also addressed issues and improved tests related to these features.
metalvulkancompilertensoropencl
ashutosh-arm/tvm
Jul 2021 - Oct 2023
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:120 pushes, 96 branches in 2 years 3 months
cpugpu-accelerationtvmdeep-learninggpu
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.