Maksim Levental
Senior ML Infrastructure Engineer at Apple
Cupertino, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMaksim Levental is a Senior ML Infrastructure Engineer in Cupertino with 12 years of experience building compiler and accelerator-aware ML tooling. He specializes in MLIR, compilers, DSLs, and backend codegen—work that spans CoreML at Apple, AMD GPU codegen for Triton and Instinct/CDNA, and eager-mode PyTorch integration via torch-mlir and IREE. Maksim’s open-source contributions include implementing an eager-mode backend for torch-mlir and performance-focused Triton changes for AMD FP8 formats, demonstrating deep practical knowledge of low-level math, mixed precision, and LLVM integration. He combines research rigor (PhD-level CS training) with production-grade engineering, repeatedly turning compiler research into deployable toolchains. Notably, his background ranges from optimizing PyTorch pipeline/memory work at Meta to enabling GPU-focused examples in SHARK Studio, revealing a pattern of bridging research, inferencing, and hardware. Colleagues rely on him to tame complex interactions between ML frameworks, code generation, and accelerator architectures.
12 years of coding experience
4 years of employment as a software developer
Bachelor of Science (BS) Mathematics, Bachelor of Science (BS) Mathematics at Florida State UniversityDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of ChicagoDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of Florida
English, Russian
Stackoverflow
1reputation
0reached
1answer
0questions
Github Skills (29)
pytorch10
python10
back-end-development10
llvm10
matrix-multiplication10
machine-learning10
mlr10
nvidia10
triton10
p810
compiler-compiler10
performance-analysis10
eager-execution10
compiler-optimization10
compiler10
Programming languages (34)
Github contributions (5)
llvm/torch-mlir
Mar 2022 - Jan 2023
The Torch-MLIR project aims to provide first class support from the PyTorch ecosystem to the MLIR ecosystem.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:118 reviews, 19 commits, 87 PRs in 10 months
Contributions summary:Maksim primarily focuses on extending the `torch-mlir` framework to support eager mode execution for PyTorch. They implemented an eager mode backend, enabling the compilation of PyTorch operations on-the-fly. The user also addresses bugs, refactors code, and enhances the framework with features like the decomposition of certain operations. They also added examples and made updates for the use of the eager mode.
pytorchmlirtorchcompilerecosystem
llvm/llvm-project
Aug 2022 - Apr 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:344 reviews, 213 PRs, 138 pushes in 2 years 7 months
Contributions summary:Maksim primarily contributed to the Python bindings for the LLVM MLIR project. Their work involved adding and modifying Python interfaces for various MLIR features, specifically the GPU dialect, LLVM dialect, and the `linalg` dialect. They implemented attribute and type casting for new classes and improved the behavior of existing functionalities such as the handling of `memref` objects. The user's changes also included adjustments to testing frameworks, test cases, and the inclusion of support for the `CLANG_CL` compiler.
compilerstechnologiesclangsubmittoolchain
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.