George Pawelczak
TPU Compiler Engineer at Google
London, England, United Kingdom
Summary
👤
Senior🎓
Top SchoolGeorge Pawelczak is a TPU compiler engineer based in London with 11 years of experience building high-performance ML compilers and runtimes for accelerators and traditional processors. He has shipped production compiler work across leading platforms—from Apple’s MLIR-based Metal stack and Graphcore’s IPU TensorFlow integration to inference portability at Wayve—and now works on the TPU compiler at Google. George combines deep systems and algorithms expertise with hands-on C++/Python backend work, contributing optimizations to widely used projects like XLA (including constant sinking and evaluator robustness). He often focuses on squeezing out performance at the edges—lowering latency and improving inference portability—while also leading teams to deliver roadmap-driven software. Pragmatic about outreach, he only considers London-based roles with local teams and prefers formal job descriptions when contacted.
10 years of coding experience
9 years of employment as a software developer
College Maths Computing Physics Further Maths and Chemistry, College Maths Computing Physics Further Maths and Chemistry at Cardinal Newman College PrestonBachelor’s Degree Computer Science, Bachelor’s Degree Computer Science at University of Bristol
Github Skills (9)
compiler-optimization10
xla10
compiler10
c-language10
cprogramming-language10
operation10
hla10
tensorflow9
testing8
Programming languages (4)
Github contributions (5)
openxla/xla
Jul 2018 - Sep 2020
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:
Back-end Developer
Back-end DeveloperContributions:35 commits in 2 years 2 months
Contributions summary:George primarily focused on improving the XLA compiler, specifically within the `hlo_evaluator.cc` and `while_loop_constant_sinking.cc` files. Their contributions involved enhancing the handling of call and fusion instructions, ensuring proper literal creation, and refactoring code to use `TF_ASSIGN_OR_RETURN`. They also implemented optimizations such as sinking constants into the conditional computation of while loops. The changes included adding new tests to ensure proper functionality.
compilercommunity-drivenmachine-learningmodular
georgepaw/tensorflow
Jun 2018 - Jul 2021
Computation using data flow graphs for scalable machine learning
Contributions:41 pushes, 39 branches in 3 years 1 month
computationscalabledata-sciencemachine-learninggraphs
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.