Sergei Lebedev is a Staff Software Engineer with 16 years of experience building low-level ML infrastructure, developer tooling, and scalable backend systems, currently working at DeepMind on JAX and Pallas. He has a strong track record contributing to high-profile open-source projects—JAX, TensorFlow, Flax, XLA and Google Research—focusing on performance, type safety, and compiler/FFI integrations that power GPU/TPU execution. His background mixes research (algorithmic bioinformatics) and hands-on engineering across CI/CD, JVM bindings, and distributed ML training, reflecting an ability to move between product-facing features and deep platform work. Sergei also improves developer experience and documentation across projects, a habit that shows in numerous Sphinx/docstring and tooling contributions. Notably, he combines systems-level changes (Pallas kernel execution, XLA type stubs) with pragmatic fixes that reduce user friction in ML libraries.
16 years of coding experience
12 years of employment as a software developer
BS, Optical Engineering, BS, Optical Engineering at ITMO University
MS, Algorithmic Bioinformatics, MS, Algorithmic Bioinformatics at St. Petersburg University of Russian Academy of Sciences
Contributions:5 reviews, 310 commits, 51 PRs in 10 years 11 months
Contributions summary:Sergei's commits primarily focus on improving the documentation of the "pyte" terminal emulator library. The changes include fixing typos in the README, updating usage examples, correcting errors in Sphinx-based documentation, and adding descriptions. These modifications enhance the clarity, accuracy, and overall quality of the documentation, making it easier for others to understand and use the library.
Static memory-efficient Trie-like structures for Python based on marisa-trie C++ library.
Role in this project:
Back-end Developer
Contributions:52 commits, 4 PRs, 52 pushes in 3 years 4 months
Contributions summary:Sergei primarily contributed to the `marisa-trie` library by implementing core functionalities and improving existing features. Specifically, they added rich comparisons and an `__iter__` method to the `_Trie` class, enhancing the library's usability. Further improvements were made by switching to low-level iteration in `__richcmp__` and incorporating NodeOrder support to improve trie comparison capabilities.
memorypythoncpp2-xtrie
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.
Request Free Trial
Sergei Lebedev - Staff Software Engineer at Google DeepMind