Kunwar Grover
Senior Software Development Engineer at AMD
City of Edinburgh, Scotland, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolKunwar Grover is a Senior Software Development Engineer based in Edinburgh with six years of experience specializing in compilers, MLIR/IREE, and GPU code generation. He contributes to high-profile open-source projects such as LLVM and IREE, focusing on bufferization, memory space management, and transform passes that improve GPU shared-memory and workgroup codegen. At AMD (and previously at nod.ai), he has driven practical compiler optimizations that surfaced non-obvious correctness fixes—like rematerialization to address dequantization bugs—bridging research ideas with production backends. His background includes academic research stints in Europe, giving him a strong foundation in polyhedral compilation and practical implementation across both runtime and compile-time systems.
6 years of coding experience
4 years of employment as a software developer
Bachelor's degree, Computer Science, Bachelor's degree, Computer Science at International Institute of Information Technology
Github Skills (13)
tensorrt10
compiler-optimization10
code-optimization10
machine-learning10
c-language10
tensor10
tensorflow10
cprogramming-language10
operation10
mlr10
llvm10
spirv9
cuda9
Programming languages (13)
Github contributions (5)
llvm/llvm-project
Jun 2021 - Mar 2025
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:269 reviews, 115 PRs, 150 pushes in 3 years 9 months
Contributions summary:Kunwar primarily worked on bufferization and memory space management within the LLVM project, specifically focusing on the MLIR component. Their contributions involved modifying code related to tensor copy insertion, memory space selection, and the implementation of the PartialReductionOpInterface for linalg.generic. They also contributed to removing dialect-specific bufferization passes and resolving buildbot failures related to these changes. The user's work suggests a focus on compiler optimization and memory management within the context of the LLVM infrastructure.
compilerstechnologiesclangsubmittoolchain
iree-org/iree
May 2023 - Apr 2025
A retargetable MLIR-based machine learning compiler and runtime toolkit.
Role in this project:ML Engineer & Compiler Developer
ML Engineer & Compiler DeveloperContributions:553 reviews, 401 PRs, 354 pushes in 1 year 10 months
Contributions summary:Kunwar contributed to the IREE compiler project, specifically focusing on machine learning and code generation aspects. Their commits added a transform operation (`transform.iree.workgroup_swizzle`) to the transform dialect for GPU code generation, enabling the optimization of workgroup memory access patterns. They also integrated and enhanced existing transform passes, such as `transform.iree.pack_shared_memory_alloc` for managing shared memory allocations, and re-enabled the `RematerializeParallelOps` pass to address a critical bug related to dequantization, ensuring that code generation functions correctly across various backends. Furthermore, they improved the raise-special-ops pass to handle tensor.extract operations and various related edge cases, indicating a focus on refining the compilation pipeline.
mlirspirvvulkantensorflowcompiler
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.