Nikita Karetnikov
Machine Learning Tech Lead at Trade Republic
Berlin, Germany
Summary
🤩
Rockstar🎓
Top SchoolNikita Karetnikov is a Machine Learning Tech Lead based in Berlin with 13 years of software engineering experience, currently leading ML efforts at Trade Republic. He combines deep expertise in PyTorch internals, JIT compilers and high-performance kernel work with hands-on systems programming in C++, CUDA, Triton and Rust. An active open-source contributor, he has helped bring up PyTorch 2 and contributed kernel and datatype fixes to the flagship pytorch/pytorch repo, and implemented AArch64 support in the symbolic execution tool Manticore. His background spans ML infrastructure, security-focused tooling and consultancy, giving him a rare blend of compiler-level insight and production ML engineering. Outside work he blends interests in math, functional programming, compilers and the demoscene, reflecting a pragmatic yet creative engineering mindset.
13 years of coding experience
5 years of employment as a software developer
Master's degree, Economics, Master's degree, Economics at Moscow State University of Geodesy and Cartography (MIIGAiK)Bachelor's degree, Computer Science, Bachelor's degree, Computer Science at Bauman Moscow State Technical University
Stackoverflow
144reputation
921reached
5answers
1question
Github Skills (25)
pytorch10
c-language10
operation10
python10
tensorrt10
binarydiff10
security10
tensorflow10
arm10
symbolic-execution10
tensor10
cprogramming-language10
emulation10
program-analysis10
testing9
Programming languages (18)
Github contributions (5)
pytorch/pytorch
Dec 2021 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer
Back-end DeveloperContributions:656 reviews, 604 commits, 299 PRs in 1 year 1 month
Contributions summary:Nikita contributed to the PyTorch library by implementing features for tensor operations. Their commits focused on supporting different data types for `addcmul` and `addcdiv` operations, as well as adding a reference implementation for `allclose`. They also addressed issues related to empty tensor behavior and made improvements to other operations, such as `movedim`, `diag_embed`, and data conversion. The user also integrated new ref implementations for mathematical functions and corrected build issues.
pythongpu-accelerationdeep-learninggpunumpy
trailofbits/manticore
Jan 2019 - Jul 2019
Symbolic execution tool
Role in this project:Back-end Developer & Security Engineer
Back-end Developer & Security EngineerContributions:17 commits, 12 PRs, 30 comments in 6 months
Contributions summary:Nikita made significant contributions to the Manticore project, focusing on adding support for the AArch64 architecture. Their work included implementing core components for the CPU, register files, and related instruction sets. Furthermore, the user addressed code quality and maintainability issues by removing x86-specific code and correcting import statements. Overall, their work enhanced Manticore's ability to analyze AArch64 binaries.
symbolic-executionbinary-analysispythonsmtethereum
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.