Serge Panev
Deep Learning Software Engineer at NVIDIA
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolSerge Panev is a Deep Learning Software Engineer with a decade of experience accelerating training and inference at NVIDIA, specializing in integrating hardware-aware optimizations into deep learning frameworks. He focuses on the intersection of GNNs, retrieval-augmented generation, and LLMs, with a practical bent toward post-training and inference optimization for real-world deployments. Serge has contributed to high-impact open-source projects—like PyTorch Geometric and NVIDIA DALI—improving test reliability, data-loading performance, and GPU-accelerated operators. Comfortable across C++, CUDA and Python, he blends low-level kernel work with Python frontend design and benchmarking to squeeze performance from DGX-class systems. Based in the Bay Area, he pairs a research curiosity about differentiable models with hands-on engineering that ships robust tooling and reproducible benchmarks.
10 years of coding experience
3 years of employment as a software developer
Bachelor of Engineering (B.E.), Computer Science, Bachelor of Engineering (B.E.), Computer Science at Griffith College CorkMaster's degree, Computer Science, Master's degree, Computer Science at EPITA: Ecole d'Ingénieurs en Informatique
French, English, Korean, Macedonian
Github Skills (31)
pytorch10
c-language10
distributed-training10
python10
image-processing10
data-engineering10
testing10
video-processing10
mxnet10
machine-learning10
horovod10
distributed-computing10
ssds10
deep-learning10
data-processing10
Programming languages (5)
Github contributions (5)
NVIDIA/DALI
Jun 2018 - Mar 2020
A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:117 commits, 176 PRs, 63 pushes in 1 year 8 months
Contributions summary:Serge primarily contributed to the development of data processing and deep learning functionality within the DALI (NVIDIA Data Loading Library) repository. Their work included implementing and refining the BoxEncoder operator, expanding support for image masks in the COCOReader, and integrating a new video reader with features such as stride and data type selection. Further, the user worked on optimizing and refactoring existing code, including improvements to the prefetching strategy and the core data structures.
pythontrainingpaddlegpu-accelerationbuilding-blocks
apache/mxnet
Jul 2019 - Oct 2021
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:32 reviews, 46 commits, 54 PRs in 2 years 4 months
Contributions summary:Serge primarily contributed to the `mxnet` deep learning framework. They implemented functionalities related to parameter management within the Gluon API, including the addition of methods for listing and resetting contexts within `ParameterDict`. Furthermore, the user factorized and optimized CUDA kernels by adding a CUDA_KERNEL_LOOP macro in multiple files. Finally, they added a mask target generator operator for Mask-RCNN, showcasing contributions to the computer vision domain.
pythonschedulerdataflowmutationdata-science
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.