Pavel Iakubovskii
AI Engineer at Praktika.ai
Lisbon, Portugal
Summary
🤩
Rockstar🎓
Top SchoolPavel Iakubovskii is an AI Engineer with 8 years of hands-on experience building production-ready computer vision systems and deep learning libraries. He is the creator and maintainer of widely used open-source segmentation tooling—packages with over 5M PyPI downloads and ~15K GitHub stars—reflecting deep expertise in semantic segmentation architectures and pretrained backbones. Pavel has applied this expertise across industry roles at Hugging Face, Denti.AI, and remote sensing startups, shipping models for medical imaging, satellite analysis, and object detection. His contributions span both research and engineering: implementing novel encoders, refactoring training pipelines, and integrating transformer and convolutional backbones into flexible APIs. Based in Lisbon, he pairs strong academic roots in deep learning for remote sensing with a talent for turning complex model architectures into reliable, user-friendly libraries.
8 years of coding experience
8 years of employment as a software developer
Master's degree Deep learning in remote sensing, Master's degree Deep learning in remote sensing at Skolkovo Institute of Science and TechnologyMaster's degree Design of technological complexes, Master's degree Design of technological complexes at Bauman Moscow State Technical University
Russian, English
Github Skills (34)
pytorch10
convolutional-neural-networks10
python10
pre-trained-model10
architecture10
image-classification10
imagenet10
machine-learning10
implement10
mask-rcnn10
keras10
image-segmentation10
densenet10
deep-learning10
tensorflow10
Programming languages (4)
Github contributions (5)
qubvel-org/segmentation_models.pytorch
Mar 2019 - Jan 2023
Semantic segmentation models with 500+ pretrained convolutional and transformer-based backbones.
Role in this project:
Back-end Developer & Algorithm Engineer
Back-end Developer & Algorithm EngineerContributions:16 releases, 159 reviews, 199 commits in 3 years 11 months
Contributions summary:Pavel primarily contributed by adding support for various VGG, DenseNet, SENet, and InceptionResNetV2 encoders to the segmentation models library. These contributions involved implementing encoder classes, loading pre-trained weights, and integrating them into the existing architecture. Furthermore, the user modified the structure of the U-Net model to support the new encoders and implemented the necessary functionality to handle the model's forward pass.
semantic-segmentationdeeplab-v3-plusunet-pytorchimage-segmentationpretrained-weights
qubvel/classification_models
May 2018 - Feb 2020
Classification models trained on ImageNet. Keras.
Role in this project:ML Engineer
ML EngineerContributions:6 releases, 99 commits, 17 PRs in 1 year 9 months
Contributions summary:Pavel primarily focused on updating and testing image classification models within the repository. Their contributions involved modifying the `weights.py` file, which likely involved updating the weights for pre-trained models such as ResNet, and creating tests in `test_imagenet.py`. These changes demonstrate their work in improving model performance and ensuring the model's accuracy for various image classification architectures. The user also addressed a bug related to image preprocessing.
mobilenetnasnetpretrained-weightstensorflowclassification
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.