Vladimir Nekrasov
Moscow, Moscow, Russian Federation
Summary
🤩
Rockstar🎓
Top SchoolVladimir Iglovikov is a CEO and machine learning/computer vision specialist with 11 years of experience building and shipping ML systems from research to production. Based in San Francisco, he founded and leads Albumentations while previously driving ML at Lyft and improving product outcomes at companies like TrueAccord and Bidgely. His background spans a PhD in physics and hands-on research—authoring Monte Carlo code and peer-reviewed papers—which informs a rigorous, data-driven approach to model design and evaluation. An active open-source contributor, he has adapted and optimized semantic segmentation models (DeepLab-ResNet, RefineNet) for real-world training and deployment, including TensorFlow conversion tooling and light-weight encoders like MobileNetV2. He often bridges academia and industry, advising startups and speaking to large data-science communities, and his work frequently focuses on pragmatic engineering decisions such as batch-norm training flags and model-weight conversion that accelerate reproducible ML. Colleagues describe him as a founder-engineer who combines deep theoretical training with practical production experience.
11 years of coding experience
Doctor of Philosophy - PhD, Computer Science, /, Doctor of Philosophy - PhD, Computer Science, / at University of AdelaideLomonosov Moscow State University
Russian, French, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (10)
semantic-segmentation10
mask-rcnn10
faster-rcnn10
pytorch10
machine-learning10
computer-vision10
deep-learning10
tensorflow10
batch-normalization10
python10
Programming languages (7)
Github contributions (5)
DrSleep/tensorflow-deeplab-resnet
Dec 2016 - Oct 2021
DeepLab-ResNet rebuilt in TensorFlow
Role in this project:
ML Engineer
ML EngineerContributions:118 commits, 5 PRs, 108 pushes in 4 years 10 months
Contributions summary:Vladimir's commits primarily involve modifying and updating the `deeplab_resnet/model.py` and related training scripts. These modifications suggest the user's focus on adapting or refining a DeepLab-ResNet model for semantic segmentation tasks. The commits specifically involve changes to batch normalization layers, including the addition of the `is_training` flag, indicating an effort to optimize the training process. Furthermore, the addition of the `npy2ckpt.py` and `convert.py` scripts reveals a focus on integrating the model with TensorFlow and converting model weights, showing a clear machine learning focus.
pascal-vocsemantic-segmentationdeeplabdeep-learningdeeplab-resnet
DrSleep/light-weight-refinenet
Oct 2018 - Jul 2020
Light-Weight RefineNet for Real-Time Semantic Segmentation
Role in this project:ML Engineer
ML EngineerContributions:46 commits, 1 PR, 38 pushes in 1 year 10 months
Contributions summary:Vladimir contributed to the training and evaluation pipeline of the light-weight RefineNet model for semantic segmentation. The commits include adding training code for the NYUv2-40 dataset, fixing broken links, and modifying the model loading process by adding `strict=False` to `load_state_dict`. Additionally, the user made modifications to notebook examples and added support for torchvision datasets and transforms. The user also added Mobilenetv2 as a possible encoder.
pytorchreal-timelight-weightsemantic-segmentationlight
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.