Vladislav Sovrasov
AI Research Engineer Scientist at Intel Corporation
Munich, Bavaria, Germany
Summary
🤩
Rockstar🎓
Top SchoolVladislav Sovrasov is an AI research engineer and scientist based in Munich with 11 years of experience specializing in deep learning, model optimization, and production-ready ML tooling at Intel. He has a strong academic foundation as a PhD student in mathematical modeling and numerical methods, and a track record of contributing practical engineering solutions such as a PyTorch FLOPs counter and integrations for the OpenVINO model zoo. His work spans research and backend implementation, including multi-camera tracking demos and rigorous test automation for OpenCV and pywinauto, reflecting a balance of reproducible science and production concerns. Colleagues know him for turning theoretical insights into robust code—improving dynamic batching, per-layer profiling, and test stability—making models faster and easier to analyze in real-world systems.
11 years of coding experience
5 years of employment as a software developer
PhD student, Mathematical modeling, numerical methods and program complexes, PhD student, Mathematical modeling, numerical methods and program complexes at State University of Nizhni Novgorod named after N.I. Lobachevsky (UNN)
German, English, Русский, Английский
Github Skills (32)
pywinauto10
uitest10
autotest10
pytorch10
opencv10
c-language10
convolutional-neural-networks10
gui-automation10
python10
testing10
machine-learning10
inference10
modello10
openvino10
automation10
Programming languages (5)
Github contributions (5)
sovrasov/flops-counter.pytorch
Aug 2018 - Nov 2022
Flops counter for convolutional networks in pytorch framework
Role in this project:
ML Engineer
ML EngineerContributions:7 reviews, 123 commits, 78 PRs in 4 years 3 months
Contributions summary:Vladislav primarily focused on developing a PyTorch-based library for counting FLOPs (floating-point operations) in convolutional neural networks. Their contributions involved implementing methods to track and compute FLOPs for various layers, including convolutions, activations, pooling, batch normalization, and RNN cells. They also added features like per-layer statistics and the ability to handle custom inputs and custom counting hooks, expanding the tool's utility and flexibility for model analysis. Furthermore, the user provided a sample script demonstrating the library's use with common PyTorch models.
flops-counterpytorchpytorch-utilsflopsdeep-learning
openvinotoolkit/open_model_zoo
Jun 2019 - Aug 2020
Pre-trained Deep Learning models and demos (high quality and extremely fast)
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:1 review, 35 commits, 13 PRs in 1 year 2 months
Contributions summary:Vladislav primarily contributed to the `smart classroom demo` within the repository, addressing issues related to dynamic batch usage and refactoring code. They fixed inconsistencies and improved code style, while also introducing a standalone function for checking dynamic batch support. Additionally, the user added the initial version of the MTMC demo, which indicates involvement in multi-camera tracking implementation. This suggests a focus on both backend functionality and model integration for the OpenVINO toolkit.
modelonnx-modelsdeep-learning-modelsmodel-zoopytorch-models
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.