Aleksei Kashapov
Software Engineer at EnCharge AI
Munich, Bavaria, Germany
Summary
🤩
Rockstar🎓
Top SchoolAleksei Kashapov is a Software Engineer and AI Research Engineer based in Munich with 10 years of experience optimizing deep learning models for production. He specializes in post-training quantization and model compression, having improved hundreds of ONNX, PyTorch, TensorFlow and OpenVINO models to achieve up to 75% size reduction and 4x inference speedups. At Intel he productized ONNX quantization pipelines, shipped subgraph matching and novel quantization-training algorithms that cut validation and fine-tuning time substantially and earned divisional recognition. His open-source contributions to the well-known OpenVINO NNCF project include implementing ONNX-based post-training quantization and hardware pattern matching improvements. Combining a physics background and CSAIL research experience in precision pose estimation, he brings a pragmatic research-to-production mindset for deploying efficient AI in robotics and edge applications.
10 years of coding experience
5 years of employment as a software developer
Master's degree, Information Systems and Technology, Master's degree, Information Systems and Technology at Skolkovo Institute of Science and TechnologyBachelor's degree, Physics, Bachelor's degree, Physics at Novosibirsk State University (NSU)
Russian, English, German
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (9)
quantization10
pytorch10
machine-learning10
deep-learning10
onnx10
model-compression10
onnxruntime9
computer-vision8
tensorflow4
Programming languages (6)
Github contributions (5)
openvinotoolkit/nncf
Sep 2020 - Jan 2023
Neural Network Compression Framework for enhanced OpenVINO™ inference
Role in this project:
ML Engineer
ML EngineerContributions:1648 reviews, 97 commits, 218 PRs in 2 years 5 months
Contributions summary:Aleksei made significant contributions to the Neural Network Compression Framework, specifically focusing on post-training quantization. Their work included implementing and refining min-max quantization, addressing bugs in export to ONNX, and enhancing hardware pattern matching. The user also developed and integrated ONNX-based post-training quantization for various model types including classification and detection. The contributions involved adding new algorithms, refining the API, and improving the framework's compatibility with ONNX models.
bertsemantic-segmentationmixed-precision-trainingtensorflowclassification
kshpv/nncf
Aug 2020 - Apr 2025
PyTorch*-based Neural Network Compression Framework for enhanced OpenVINO™ inference
Contributions:2 reviews, 5 PRs, 1527 pushes in 4 years 7 months
pytorchneural-network-compressiondeep-learningcompressioninference
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.