Frédéric Bastien
Distinguished Engineer at NVIDIA
Brossard, Quebec, Canada
Summary
🤩
Rockstar🎓
Top SchoolFrédéric Bastien is a distinguished engineer with 18 years of experience in high-performance ML infrastructure, currently leading work on the JAX team at NVIDIA. He combines deep compiler and GPU expertise—evidenced by contributions to XLA, NVTX instrumentation in TensorFlow, and low-level optimizations in Theano and DeepRec—with hands-on improvements to widely used scientific libraries like NumPy. At MILA and Université de Montréal he built and led research software teams, bridging research code and production-grade systems. His open-source footprint spans performance engineering, CI/CD and profiling for projects such as JAX, TensorFlow, Theano and convnet benchmarks, demonstrating a knack for squeezing performance out of convolution and array kernels. Notably, he fixed subtle numerical and float32 issues across libraries and improved CI reliability for JAX’s nightly builds—work that often surfaces only under heavy GPU/TPU workloads. He holds a Master’s in Computer Science from Université de Montréal and is based in Brossard, Quebec.
18 years of coding experience
9 years of employment as a software developer
DEC, Science Pur, DEC, Science Pur at Collège de MaisonneuveMaster's degree, Computer Science, Master's degree, Computer Science at Université de Montréal
Github Skills (46)
code-optimization10
debug10
github-ci10
benchmark10
performance-monitor10
c-language10
convolutional-neural-networks10
performance-analytics10
compilation10
python10
machine-learning10
benchmarking10
performance-measurement10
numpy10
performance-analysis10
Programming languages (12)
Github contributions (5)
lisa-lab/pylearn2
May 2008 - Aug 2021
Warning: This project does not have any current developer. See bellow.
Role in this project:
ML Engineer
ML EngineerContributions:3985 commits, 23 PRs, 15 pushes in 13 years 5 months
Contributions summary:Frédéric contributed to the pylearn2 library by fixing a crash when compiling an error function within the experiment.py file. They also added names to Theano functions and fixed autoencoder tests, specifically addressing issues in float32. Furthermore, the user corrected tests to work correctly in the presence of float32. In addition, the user implemented a MaxPool op that wrap Alex's maxpool code.
javascripttypescript
Theano/Theano
May 2012 - Dec 2022
Theano was a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. It is being continued as PyTensor: www.github.com/pymc-devs/pytensor
Role in this project:Backend Developer & Performance Engineer
Backend Developer & Performance EngineerContributions:1 review, 5435 commits, 1765 PRs in 10 years 8 months
Contributions summary:Frédéric primarily focused on optimizing and implementing performance-critical code within the Theano library, a project centered around efficient mathematical expression evaluation involving multi-dimensional arrays. Their contributions included moving specific tests for softmax gradients to the test_dnn.py file and removing unused parameters to streamline the code. They also made various modifications to address compilation errors, particularly those related to C++ code and mixed data types, demonstrating a solid understanding of the library's internal workings and optimization techniques.
python-librarymathmulti-dimensionalpythonevaluate
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.