David Tanner
PMTS Software Development Engineer at AMD
Apex, North Carolina, United States
Summary
🤩
Rockstar🎓
Top SchoolDavid Tanner is a performance-focused software engineer with 13 years of experience optimizing accelerator and system-level ML performance, currently working at AMD after leading hardware/software co-design for TPU SuperPods at Google. He helped push GEMM and dense compute efficiency forward at AMD and NVIDIA—creating Tensile and contributing to rocBLAS/clBLAS and cuBLAS/cuDNN—and built compiler passes that auto-detect and optimize tensor contractions. His work spans low-level assembly kernel generation, autotuning frameworks, and compiler backend enhancements, delivering near peak hardware efficiency for real-world ML workloads. At Google he bridged networking, compiler, and chip teams to model perf/TCO with tools like TfSim, showing his strength in turning cross-disciplinary research into deployable platform designs. Fluent in Albanian and grounded in a PhD in Biophysics and Computational Biology, he brings a rare combination of deep scientific training and practical high-performance systems engineering.
13 years of coding experience
19 years of employment as a software developer
Highschool General, Highschool General at GilmanDoctor of Philosophy (Ph.D.) Biophysics and Computational Biology, Doctor of Philosophy (Ph.D.) Biophysics and Computational Biology at University of Illinois Urbana-ChampaignHighschool Physics Chemistry Biology Math, Highschool Physics Chemistry Biology Math at WaterfordBachelor’s Degree Theoretical and Mathematical Physics, Bachelor’s Degree Theoretical and Mathematical Physics at Brigham Young University
English, Albanian
Stackoverflow
559reputation
63kreached
9answers
2questions
Github Skills (20)
opencl10
c-language10
matrix10
blas10
mat10
performance-optimization10
cprogramming-language10
linear-algebra10
maven6
mojo6
junit6
encoding6
xargs6
pomxml6
java6
Programming languages (4)
Github contributions (5)
ROCm/rocBLAS
Oct 2015 - Dec 2017
Next generation BLAS implementation for ROCm platform
Role in this project:
Back-end Developer
Back-end DeveloperContributions:97 commits, 40 PRs, 16 pushes in 2 years 2 months
Contributions summary:David's commits primarily focused on defining data structures and functions for the rocBLAS library, specifically related to matrix operations. They implemented data types for matrices and scalars, and defined matrix operations, including matrix multiplication (gemm). The commits also involved the inclusion of headers and the integration of Cobalt (later Tensile), suggesting the user was working on incorporating optimized BLAS functionality.
blasnext-generationrocmhip
clMathLibraries/clBLAS
Jun 2015 - May 2017
a software library containing BLAS functions written in OpenCL
Role in this project:Back-end Developer / Performance Engineer
Back-end Developer / Performance EngineerContributions:35 commits, 27 PRs, 13 pushes in 1 year 10 months
Contributions summary:David primarily contributed to the clBLAS library by implementing and optimizing BLAS (Basic Linear Algebra Subprograms) functions, specifically focusing on the ZGEMM (complex double-precision matrix multiplication) kernel for AMD's "Hawaii" architecture. Their work involved modifying OpenCL kernel templates, fixing bugs, and enabling performance-related features like API call counts within the client code. Further contributions included adding performance data and fixing scripts related to graph generation for performance analysis, with adjustments and fixes to the build process as well.
cudalibrary-softwaregpublasopencl
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.