Andy Adinets
Senior AI Developer Technology Engineer at NVIDIA
Munich, Bavaria, Germany
Summary
🤩
Rockstar🎓
Top SchoolAndy Adinets is a Senior AI Developer Technology Engineer based in Munich with 14 years of experience building high-performance ML and GPU software. At NVIDIA since 2017 he focuses on low-level optimization and GPU-enabled ML primitives, contributing production-grade improvements to projects like XGBoost and cuML that accelerate distributed gradient boosting and RAPIDS workflows. His background spans research and SRE roles at Google and Forschungszentrum Jülich, giving him a strong combination of systems reliability, parallel algorithm design, and scientific computing. Notably, his open-source work includes multi-GPU support, GPU quantile algorithms, and faster radix/sort stability fixes in NVIDIA CUB—efforts that materially improve performance for large-scale ML workloads.
14 years of coding experience
9 years of employment as a software developer
Moscow State UniversityLomonosov Moscow State University
Github Skills (19)
gbm10
algorithms10
xgboost10
c-language10
gpgpu10
gpu-programming10
radix-sort10
machine-learning10
distributed-systems10
kernel10
cub10
parallel-computing10
gpu10
performance-optimization10
cubit10
Programming languages (4)
Github contributions (5)
rapidsai/cuml
Jun 2019 - Aug 2022
cuML - RAPIDS Machine Learning Library
Role in this project:
ML Engineer
ML EngineerContributions:212 reviews, 190 commits, 50 PRs in 3 years 3 months
Contributions summary:Andy implemented and refined core functionalities within the cuML library, a RAPIDS Machine Learning Library. Their contributions focused on low-level optimization and building kernels. Specifically, they introduced new functions for atomic operations (min/max bits) within the `stats/minmax.h` file. They also consolidated code with the existing minmax functions and added tests to verify those features.
cudacumlnvidiadata-sciencegpu
NVIDIA/cub
Dec 2019 - Nov 2022
[ARCHIVED] Cooperative primitives for CUDA C++. See https://github.com/NVIDIA/cccl
Role in this project:Back-end Developer / Performance Engineer
Back-end Developer / Performance EngineerContributions:43 reviews, 18 commits, 12 PRs in 2 years 11 months
Contributions summary:Andy's commits primarily focus on optimizing the `cub` library, a cooperative primitives library for CUDA. Their work includes implementing a faster radix sort algorithm and improving the stability of sorting for floating-point numbers. They also addressed various performance bottlenecks, fixed overflows, and optimized aspects of the upsweep/downsweep sorting algorithms. Several commits are related to improving the underlying algorithm.
cxxprimitivesnvidia-hpc-sdknvidiacpp20
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.