Sergio Colmenarejo
Member Of Technical Staff at Microsoft AI
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolSergio Colmenarejo is a machine learning engineer with 12 years' experience building research-grade ML systems and core numerical libraries, currently a Member of Technical Staff at Microsoft AI after a long tenure as Senior Staff Research Engineer at DeepMind. He has deep hands-on expertise in reinforcement learning and tensor libraries, contributing to flagship projects like DeepMind's Acme and the Torch core (torch7/nn), including offline RL agents, recurrent-state tweaks, and low-level tensor operations and memory optimizations. Trained with distinction in Computational Statistics and Machine Learning at UCL and grounded in computer science and math from Universidad Autónoma de Madrid, he bridges rigorous research and production engineering. Colmenarejo’s work shows a pattern of upgrading and modernizing legacy ML codebases (TensorFlow/Sonnet migrations) and adding robust test coverage—skills that make him effective at keeping research code run-ready at scale. An interesting thread through his career is moving between deep research problems and gritty backend improvements, from bioinference startups to world-class ML labs.
12 years of coding experience
13 years of employment as a software developer
Computer Science and Mathematics, Computer Science and Mathematics at Universidad Autónoma de MadridMaster of Science (MSc), Computational Statistics and Machine Learning, Distinction, Master of Science (MSc), Computational Statistics and Machine Learning, Distinction at University College London
Spanish, English
Github Skills (22)
pytorch10
numerical10
computation10
operation10
python10
tensorrt10
machine-learning10
c1110
reinforcement-learning10
c1710
lua10
numerics10
deep-learning10
tensorflow10
neural-network10
Programming languages (5)
Github contributions (5)
google-deepmind/learning-to-learn
Dec 2016 - Jul 2017
Learning to Learn in TensorFlow
Role in this project:
ML Engineer
ML EngineerContributions:11 commits, 5 PRs, 10 pushes in 7 months
Contributions summary:Sergio primarily focused on adapting the codebase to be compatible with newer versions of TensorFlow (0.12 and later) by updating deprecated functions and initializers. They made changes to various test files, ensuring tests passed after the TensorFlow upgrades. Additionally, the user refactored code to use Sonnet, demonstrating a focus on maintaining and improving the codebase's compatibility with evolving ML frameworks.
deep-learningneural-networkslearning-to-learntensorflowartificial-intelligence
torch/nn
Jun 2014 - Feb 2016
Role in this project:Back-end Developer
Back-end DeveloperContributions:9 commits, 3 PRs in 1 year 8 months
Contributions summary:Sergio significantly contributed to the `nn` module within the `torch` repository, primarily focused on improving the functionality and usability of table-based operations. They added support for minibatch processing to `JoinTable` and `SplitTable` modules, enabling these modules to handle tensors with varying numbers of dimensions. Further improvements included refactoring the modules by removing the `setNumInputDims` method and introducing the `View` module for creating new views of tensors. Additionally, they addressed memory management issues and added support for negative indices in `SplitTable`.
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.