Benjamin Chetioui
Staff Software Engineer at Google DeepMind
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolBenjamin Chetioui is a software engineer and PhD candidate in Programming Languages at the University of Bergen with 11 years of experience building compilers and high-performance ML infrastructure. He has driven block-level codegen and custom kernel infrastructure at Google (XLA) and currently contributes to JAX/Pallas at DeepMind, focusing on GPU code generation and performance-critical kernel emitters. An active open-source contributor, his work spans LLVM/MLIR, TensorFlow/XLA, and JAX-to-TensorFlow conversion, improving lin‑alg primitives and buffer semantics for efficient tensor computation. Beyond compilers and ML, he brings deep expertise in information security and cryptography—co-founding a top-ranked CTF team and publishing peer-reviewed work in post-quantum cryptography. He combines formal foundations (Mathematics of Arrays, generic programming, Magnolia language) with hands-on systems engineering, making him fluent at turning mathematical ideas into practical, high-performance implementations.
12 years of coding experience
4 years of employment as a software developer
Master’s Degree, Computer Software Engineering, Master’s Degree, Computer Software Engineering at University of Strasbourg
English, French, Norwegian
Stackoverflow
50reputation
20kreached
10answers
7questions
Github Skills (31)
conversions10
c-language10
python10
data-manipulation10
gpu-programming10
machine-learning10
mlr10
triton10
tensorflow10
bazel10
compiler10
xla10
jax10
cprogramming-language10
linear-algebra10
Programming languages (9)
Github contributions (5)
openxla/xla
Aug 2020 - Jan 2023
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:
Back-end Developer
Back-end DeveloperContributions:184 reviews, 32 commits, 286 comments in 2 years 5 months
Contributions summary:Benjamin's commits primarily focus on enhancing the XLA compiler, particularly the implementation of the `scatter` operation and the integration of buffer semantics. The contributions include fixing documentation, improving the compatibility of THLO ops with buffer semantics, and adding a sorting operation to the dialect. The changes also involve adding tests and fixing issues related to operations like `concatenate` and adding conversion paths for HLO operations.
compilermachine-learning
jax-ml/jax
Nov 2020 - Jan 2021
Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more
Role in this project:Back-end Developer & MLOps Engineer
Back-end Developer & MLOps EngineerContributions:183 reviews, 34 commits, 122 PRs in 1 month
Contributions summary:Benjamin primarily worked on adding support for features related to the conversion of JAX code to TensorFlow, specifically focusing on operations related to linear algebra (e.g., triangular solve, eigen decomposition) and array manipulation. They added and improved the jax2tf conversion code for various primitives, including `top_k`, `scatter`, `reduce_window`, and `conv_general_dilated`. Furthermore, the user addressed bugs and inconsistencies related to data type handling and compatibility between JAX and TensorFlow, indicating a focus on MLOps.
gpujittpujax
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.