Derek Murray
Principal Software Engineer at Google DeepMind
Redwood City, California, United States
Summary
🤩
Rockstar🎓
Top SchoolDerek Murray is a Senior Staff Software Engineer at Google in Redwood City with 18 years of experience and a PhD from the University of Cambridge. He specializes in ML systems and TensorFlow internals, turning research models into production-grade, high-performance implementations. An active open-source contributor to tensorflow, google-research and DeepRec, he has improved data pipelines, migrated deprecated TF APIs, and optimized core tensor operations and execution engine locking to boost efficiency. Derek is known for digging into low-level bottlenecks—e.g., ConvertToCooTensorOp, FrameState::GetIteration, and scalar tensor wrapping—so research code runs reliably at scale.
19 years of coding experience
15 years of employment as a software developer
MSc, MSc at The University of EdinburghPhD, PhD at University of CambridgeBSc (Hons), BSc (Hons) at University of Glasgow
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (20)
c-language10
data-pipelines10
python10
machine-learning10
tpu10
data-preprocessing10
deeplearning-ai10
deep-learning10
tensorflow10
performance-optimization10
scalability10
cprogramming-language10
data-pipeline10
optimization10
distributed-training9
Programming languages (9)
Github contributions (5)
tensorflow/tensorflow
Dec 2015 - May 2020
An Open Source Machine Learning Framework for Everyone
Role in this project:
Back-end Developer
Back-end DeveloperContributions:4 reviews, 1170 commits, 189 PRs in 4 years 5 months
Contributions summary:Derek focused on optimizing the performance of the `ConvertToCooTensorOp` operation, addressing computational and allocation issues. They implemented several improvements, including conditional gain-rescaling avoidance and efficient handling of scalar weights. Furthermore, the user made improvements in the code for extracting row IDs from SparseTensor. These changes demonstrate a strong understanding of performance optimization techniques within the TensorFlow framework.
pythondata-sciencedeep-learningmlmachine-learning
DeepRec-AI/DeepRec
Sep 2019 - Feb 2020
DeepRec is a high-performance recommendation deep learning framework based on TensorFlow. It is hosted in incubation in LF AI & Data Foundation.
Role in this project:Back-end Developer
Back-end DeveloperContributions:13 commits in 5 months
Contributions summary:Derek primarily focused on optimizing and refactoring core components of the deep learning framework, specifically related to tensor operations and execution efficiency. They optimized the tensor-wrapping of scalar values within the CPU, making it more efficient. Furthermore, the user improved performance in the execution engine by optimizing the FrameState::GetIteration() function, which reduces the time spent under the lock within ExecutorState::PropagateOutputs(). These changes indicate an effort to improve performance in the framework.
pythondeep-learningrecommendationdistributed-trainingmachine-learning
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.