Ivan Lazarevich
Member Of Technical Staff at Cerebras Systems
Old Toronto, Ontario, Canada
Summary
🤩
Rockstar🎓
Top SchoolIvan Lazarevich is a Member of Technical Staff with 11 years of experience applying research-grade deep learning to real-world inference and compression problems. He has driven hardware-aware model optimization and NAS at Deeplite and contributed core compression features to Intel’s OpenVINO tooling, including BN adaptation and Mask-RCNN integrations in the popular NNCF repository. At Cerebras he focuses on LLM inference optimization—speculative decoding, long-context handling, RAG and genomics applications—bridging cutting-edge research with production inference. Trained as a computational neuroscientist (PhD-level work at École normale supérieure), he brings a strong theoretical grounding to practical performance engineering and a track record of shipping algorithmic innovations for constrained hardware.
10 years of coding experience
9 years of employment as a software developer
Doctor of Philosophy - PhD, Computational Neuroscience, Doctor of Philosophy - PhD, Computational Neuroscience at Ecole normale supérieure
Английский
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (14)
compression10
quantization10
sparse10
pytorch10
deeplearning-ai10
deep-learning10
batch-normalization10
onnx7
tensorflow7
object-detection7
bert5
openvino5
nlp5
transformers5
Programming languages (2)
Github contributions (5)
openvinotoolkit/nncf
Jul 2020 - Sep 2021
Neural Network Compression Framework for enhanced OpenVINO™ inference
Role in this project:
ML Engineer
ML EngineerContributions:92 reviews, 20 commits, 22 PRs in 1 year 1 month
Contributions summary:Ivan primarily contributes to the Neural Network Compression Framework (NNCF), focusing on implementing and improving algorithms for model compression. Their work includes enabling and refining batch normalization (BN) adaptation techniques within the quantization process, particularly in conjunction with filter pruning and sparsity methods. They extend the configuration schema to accommodate BN adaptation parameters and address issues in the model export process. The user also added configurations for integrating Mask-RCNN with the framework.
bertsemantic-segmentationmixed-precision-trainingtensorflowclassification
lzrvch/vanyalzr.github.io
Feb 2018 - Feb 2025
Contributions:50 pushes, 1 branch in 7 years
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.