Babak Poursartip
SMTS - Software Development Engineer at AMD
Austin, Texas, United States
Summary
👤
Senior🎓
Top SchoolBabak Poursartip is a computational scientist and software engineer with 15+ years solving PDEs and nine years of professional software experience, now building high-performance numerical libraries and ML infrastructure at AMD. He combines deep domain expertise in finite element/finite volume methods, seismic and acoustic wave simulation, and inverse problems with hands-on HPC skills in C++, CUDA/HIP, PETSc, MPI and OpenMP. His work spans research-grade modeling—mesh generation, PMLs, fluid-structure interaction—and production performance engineering, contributing to prominent ROCm projects like rocBLAS and Tensile to optimize high-precision GEMM kernels. Academically grounded with a PhD in Computational Mechanics and teaching experience at UT Austin, he translates theory into scalable parallel codes and large-scale visualization pipelines using VTK/HDF5. Notably, he has a track record of end-to-end implementations from 5k-line finite-element solvers to GPU backend optimizations, reflecting both deep numerical insight and practical performance tuning. Based in Austin, he maintains an open-source portfolio that surfaces his blend of research rigor and production-oriented engineering.
9 years of coding experience
11 years of employment as a software developer
Doctor of Philosophy - PhD, Computational Mechanics, 4.00 GPA, Doctor of Philosophy - PhD, Computational Mechanics, 4.00 GPA at The University of Texas at AustinAmirkabir University of TechnologyBachelor of Science, ENGINEERING, Bachelor of Science, ENGINEERING at Shahid Chamran UniversityDimlopa, Math. and Physics, Dimlopa, Math. and Physics at National Organization for Development of Exceptional Talents
Persian, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (18)
assembly10
c-language10
blas10
hip10
gemfire10
performance-optimization10
cprogramming-language10
python9
opencl9
tensorrt9
contract9
gpu9
tensor9
contraction9
machine-learning8
Programming languages (5)
Github contributions (5)
ROCm/rocBLAS
Dec 2021 - Jan 2023
[DEPRECATED] Moved to ROCm/rocm-libraries repo
Role in this project:
Back-end Developer / Performance Engineer
Back-end Developer / Performance EngineerContributions:14 reviews, 29 commits, 19 PRs in 1 year 1 month
Contributions summary:Babak's contributions primarily focused on optimizing and enhancing the rocBLAS library, specifically concerning High Precision Accumulate (HPA) GEMM kernels. They modified the code to normalize naming conventions for HPA kernels and implemented memory allocation strategies for GSU (Generalized Solution Update) operations. Furthermore, the user added test cases for updated HPA functionality and updated the documentation, indicating a focus on performance improvements and code maintainability. The commits also include updates to code object versions and library logic, indicating work to keep the library current.
blasnext-generationrocmhip
ROCm/Tensile
Dec 2021 - Dec 2022
[DEPRECATED] Moved to ROCm/rocm-libraries repo
Role in this project:Back-end Developer
Back-end DeveloperContributions:151 reviews, 16 commits, 122 PRs in 11 months
Contributions summary:Babak contributed to the `rocm/tensile` repository by addressing a bug related to High-precision fp16 (HBH) in the kernel assembly and modifying compute data types. They also normalized naming conventions for various HPA types and updated the GSU (Global Split U) functionality, including adding a SingleBuffer algorithm for HSS/BSS gemm functions. Additionally, the user enhanced the efficiency script by including all gemm data types and added zero to valid values for StaggerUStride. Finally, the user updated the library version.
amdpythontensorhipauto-tuning
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.