Kimish Patel
Senior Staff Software Engineer at Google
San Jose, California, United States
Summary
🤩
Rockstar🎓
Top SchoolKimish Patel is a Senior Staff Software Engineer in San Jose with nine years of experience driving on-device machine learning across Meta’s AR products, specializing in computer vision, speech and language models. He bridges semiconductor and software domains—leveraging a PhD in electrical engineering and hands-on CPU/GPU microarchitecture experience—to co-design models and optimize ML runtimes for power- and latency-constrained devices. At Meta he’s led performance tooling, debugging, and deployment efforts; his open-source contributions to high-profile projects like Apache TVM and PyTorch include tensorization, AVX2 GEMM optimizations, and on-device quantization work that tangibly improve inference performance. Known for pragmatic hardware-software codesign, he combines low-level architecture insight with production ML engineering to push real-time AR experiences on edge hardware.
9 years of coding experience
15 years of employment as a software developer
Doctor of Philosophy (PhD) Electrical Engineering, Doctor of Philosophy (PhD) Electrical Engineering at University of Southern CaliforniaBachelor's degree Information Technology, Bachelor's degree Information Technology at Gujarat UniversityMachine Learning Nanodegree, Machine Learning Nanodegree at UdacityVisiting Researcher, Visiting Researcher at Politecnico di Torino
English, Hindi, Gujarati, Italian
Github Skills (19)
tvm10
pytorch10
c-language10
performance-monitor10
python10
avx10
machine-learning10
deeplearning-ai10
compiler-compiler10
performance-analysis10
deep-learning10
performance-optimization10
compiler10
quantization10
cprogramming-language10
Programming languages (6)
Github contributions (5)
pytorch/pytorch
Jan 2019 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:918 reviews, 395 commits, 345 PRs in 4 years
Contributions summary:Kimish primarily contributed to the PyTorch library, focusing on deep learning and machine learning related tasks. The commits involved implementing and modifying components related to quantization, including observer insertion, the creation of quantized operations and the development of a wrapper API for on-device quantization. The user also worked on thread safety for xnnpack-based convolution operations. Furthermore, they were involved in the generation and embedding of shader parameters for Vulkan, optimizing performance and fixing Vulkan profiling issues.
gpu-accelerationneural-networkpythonautogradgpu
apache/tvm
Sep 2019 - Nov 2019
Open Machine Learning Compiler Framework
Role in this project:ML Engineer
ML EngineerContributions:5 commits, 7 PRs, 52 comments in 1 month
Contributions summary:Kimish primarily contributed to the TVM compiler stack, focusing on enabling and optimizing tensorization techniques for various hardware backends. Their work included modifying the tensorize functionality, addressing bugs, and integrating AVX2-based GEMM optimizations. They also exposed the `llvm.nearbyint` intrinsic and made modifications to internal compiler passes, particularly related to storage flattening and loop partitioning, to improve performance and correctness.
compilermachine-learningtensordeep-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.