Mahesh Ravishankar
Distinguished Compiler Engineer (AI) at NVIDIA
Redmond, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolMahesh Ravishankar is a senior software development manager with 9 years of industry experience leading compiler and ML infrastructure work across AMD, Google, and NVIDIA. He specializes in MLIR/LLVM-based compiler backends and loop/tiling optimizations, with notable open-source contributions to MLIR (tensorflow/mlir) and the LLVM project that improved subview semantics, loop coalescing, and reduction interfaces. At AMD he leads the Shark/IREE team focused on taking ML frameworks from front-end integration through device-specific codegen and performance tuning for CPUs and GPUs. His background blends deep research (PhD in CS&E) on inspector-executor transformations for irregular computations with production engineering on GPU compilers and DSLs for high-performance workloads. Colleagues benefit from his rare combination of compiler theory, practical backend engineering, and experience shipping GPU-targeted optimizations at scale. He is based in Seattle and actively recruits and mentors engineers into fast-growing ML compiler teams.
10 years of coding experience
12 years of employment as a software developer
Indian Institute of Technology MadrasDoctor of Philosophy, Computer Science and Engineering, Doctor of Philosophy, Computer Science and Engineering at The Ohio State University
English, Hindi, Kannada
Github Skills (6)
compiler-optimization10
c-language10
cprogramming-language10
mlr10
spirv9
tensorflow8
Programming languages (7)
Github contributions (5)
llvm/llvm-project
Dec 2019 - May 2026
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:783 reviews, 103 PRs, 281 pushes in 6 years 5 months
Contributions summary:Mahesh made significant contributions to the MLIR project, focusing on the SCF dialect. They modernized and refactored the `coalesceLoops` method to handle `scf.for` loops with iter_args and also made changes to avoid generating unnecessary division/remainder operations. Further contributions included refactoring the `PartialReductionOpInterface` to align with the `TilingInterface` and adding a utility for moving operation dependencies. The user's work involved changes to the core compiler infrastructure, with a focus on loop transformations and optimization within the MLIR framework.
compilerllvmtoolchain
tensorflow/mlir
Nov 2019 - Dec 2019
"Multi-Level Intermediate Representation" Compiler Infrastructure
Role in this project:Back-end Developer
Back-end DeveloperContributions:20 commits, 2 PRs, 90 comments in 27 days
Contributions summary:Mahesh primarily focused on modifying and improving the `SubViewOp` within the MLIR compiler infrastructure. Their contributions involved refining the specification of the `SubViewOp`, including the handling of static and dynamic offsets, sizes, and strides, and adding canonicalization patterns. Furthermore, the user added verification checks to ensure the result type of the subview operation is consistent with which parameters are static or dynamic. They also made changes to support casting of memrefs with static strides to dynamic strides.
compiler
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.