Onur Yilmaz
Machine Learning Engineer at NVIDIA
Summary
🤩
RockstarOnur Yilmaz is a Machine Learning Engineer with seven years of experience specializing in GPU-accelerated recommender systems and production ML pipelines. He has hands-on expertise optimizing high-performance CUDA code and integrating GPU linear-algebra primitives into frameworks like NVIDIA HugeCTR and cuML to squeeze out inference and training performance. Onur has also worked on NVTabular to bridge feature-engineering workflows with Hugectr and Triton inference, demonstrating an emphasis on deployable, production-grade ML. His contributions span backend development, MLOps, and numerical correctness fixes—e.g., improving explained-variance in cuML PCA—highlighting both systems-level thinking and attention to algorithmic detail. Based in Turkey, he brings a pragmatic focus on shipping scalable, GPU-first solutions for CTR and tabular ML use cases.
7 years of coding experience
Github Skills (21)
algorithm10
python10
machine-learning10
pca10
machine-learning-algorithms10
feature-engineering10
deeplearning-ai10
mlops10
deep-learning10
gpu10
optimisation10
recommender-system10
cuda10
cpp10
cuml10
Programming languages (4)
Github contributions (5)
rapidsai/cuml
Oct 2018 - Jan 2020
cuML - RAPIDS Machine Learning Library
Role in this project:
ML Engineer
ML EngineerContributions:211 commits, 32 PRs, 12 pushes in 1 year 3 months
Contributions summary:Onur fixed bugs and enhanced functionality within the notebooks, specifically related to the truncated SVD and PCA implementations within the cuML library. These changes involved correcting issues in the explained variance calculations and improving the performance of the implementations. Further modifications were introduced for the recently added KMeans.
cudacumlnvidiadata-sciencegpu
NVIDIA-Merlin/NVTabular
Apr 2020 - Sep 2021
NVTabular is a feature engineering and preprocessing library for tabular data designed to quickly and easily manipulate terabyte scale datasets used to train deep learning based recommender systems.
Role in this project:Back-end Developer & MLOps Engineer
Back-end Developer & MLOps EngineerContributions:38 reviews, 46 commits, 28 PRs in 1 year 4 months
Contributions summary:Onur primarily contributed to integrating and updating the Hugectr output format within the NVTabular workflow. Their work involved modifying the `io.py` and `workflow.py` files to support the Hugectr integration, including defining the data format and generating outputs compatible with the Hugectr framework. The user also initiated the implementation of Triton Inference for the NVTabular models and contributed to supporting the PyTorch backend within the NVTabular inference pipeline. These changes demonstrate an effort to prepare NVTabular workflows for deployment within a production machine-learning environment.
tsneengineeringtensorflowpreprocessingnvidia
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.