Jonathan Dekhtiar
Senior Engineer - WheelNext Lead at NVIDIA
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolJonathan Dekhtiar is a Senior Engineer and WheelNext lead at NVIDIA with 11 years of experience building and optimizing deep learning systems for production. With a PhD-focused background in deep learning and applied math, he blends research rigor with hands-on engineering across TensorFlow, TensorRT, PyTorch and distributed training (Horovod), and has driven performance wins for low-latency inference. He is an active open-source contributor to prominent projects like TensorLayer, TensorFlow-TensorRT and NeMo, where his work spans API improvements, GPU prefetching fixes, AMP integration and benchmarking for TF-TRT. Based in San Francisco, he pairs systems-level expertise (C/C++, Docker, cloud) with data-science tooling in Python, and has a knack for smoothing build/CI friction—evident from NeMo install automation. Colleagues describe him as a pragmatic engineer who moves experimental models into high-performance production.
11 years of coding experience
7 years of employment as a software developer
Erasmus, Computer Science, Erasmus, Computer Science at Hamburg University of TechnologyDoctor of Philosophy - PhD, Deep Learning, Doctor of Philosophy - PhD, Deep Learning at Université de Technologie de Compiègne (UTC)
English, German, French
Stackoverflow
3,486reputation
287kreached
75answers
31questions
tensorflow
top-5%
Github Skills (40)
dependency-management10
debug10
benchmark10
distributed-training10
python10
tensorrt10
machine-learning10
benchmarking10
inference10
horovod10
reinforcement-learning10
deeplearning-ai10
deep-learning10
tensorflow10
performance-optimization10
Programming languages (24)
Github contributions (5)
tensorflow/tensorrt
Dec 2020 - Jan 2023
TensorFlow/TensorRT integration
Role in this project:
ML Engineer & Performance Engineer
ML Engineer & Performance EngineerContributions:21 reviews, 159 commits, 31 PRs in 2 years 1 month
Contributions summary:Jonathan implemented and benchmarked the integration of TensorFlow-TensorRT (TF-TRT) for low-latency inference within the TensorFlow/TensorRT repository. Their contributions primarily focused on creating and running scripts to evaluate model performance, including measuring average step times and throughput using TF-TRT in conjunction with various precisions. The user optimized inference by leveraging TF-TRT's capabilities across different precisions and batch sizes, with the goal of accelerating model execution.
nvidia-dockerdeep-learningtensorflowtensorrt
tensorlayer/TensorLayer
Jan 2018 - Jan 2019
Deep Learning and Reinforcement Learning Library for Scientists and Engineers
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:13 releases, 472 commits, 195 PRs in 11 months
Contributions summary:Jonathan primarily contributed to the development and maintenance of the TensorLayer library, focusing on enhancing the layer API and integrating new features. Their work involved updating existing layer functionalities, addressing issues, and integrating new functionalities such as support for features. The contributions included modifications to the core codebase, documentation updates, and integration of features. This indicates a strong focus on improving the library's capabilities.
tensorflow-tutorialscientistspythongoogledeep-reinforcement-learning
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.