Мой круг становится Хабр Карьерой! Узнать подробности
Обо мне

Paul Birulin, email pavel.birulin@gmail.com, What’sUp and cell +1-416-605-7858, +7-938-457-2791, looking for the position of Machine/Deep Learning Engineer 

  • Proficient in Machine/Deep Learning (Supervised, Unsupervised, Weakly/Semi supervised learning, Regression/Classification/Clustering), Deep Neural Nets (  AutoEncoders/GANs/LSTM/Object Detectors), SVM, predictive models
  • Proficient in Keras/Tensoflow, Python, numpy, scikit, Matlab, Caffe, OpenCV, Matplotlib, CUDA, SQL, Tableau, Azure, experience in C++ as well, proficient in Computer Vision
  • Proficient in object-oriented design, architecture and development of production-level software systems
  • Good experience in Kubernetes/Docker, IoT, CI/CD pipeline, SQA
  • Strong focus and attention on business and clients needs and requirements, agile business-driven approach for system development, on time and on budget
Профессиональные навыки
Опыт работы
Российский Исследовательский Центр Samsung
Sr. Staff Engineer
Март 2018 — Сентябрь 2019 (1 год и 7 месяцев)

Projects: Development various ML models, datasets, training methods, loss functions and architectures.         

Training models, fitting hyper-parameters etc (Python, Tensorflow, Keras, MXnet, PyTorch,  Caffe):

  • - Image re-draw/enhancing of images with lost details/features, patent made and submitted, paper is in progress
  • - Image quality classification
  • - Depth prediction (map of distances) from single RGB image for Indoor and Outdoor for robotics applications with strict requirements for performance, accuracy and hardware; top KITTI completion results were achieved with much lighter architecture dedicated for smartphones and Intel Movidius Neural stick. Patent and paper is in progress.
  • - Other projects include smartphone glass inspection for scratches, image segmentation, deblur.
Mexia Interactive Inc
Sr. Software Engineer and architect of ML systems
Сентябрь 2014 — Декабрь 2017 (3 года и 4 месяца)
  • - Development of the ML based IoT systems for people and objects recognition, counting and tracking, post-analysis in airports and malls from WiFi and video sensors on CPU&GPU servers using Neural Networks (Python, PyCuda, Matlab, Caffe, numpy, OpenCV, SQL, Azure, Kubernetes, Docker). The system is running in ten large airports around the word (LaGuardia, Changi, Belfast, Liverpool, Larnaka/Pafos etc) processing data from hundreds cameras in real-time.
  • - Development of the Neural Networks architectures, setup and training: for headcount in dense crowds on images and real-time video, for linecount on CPU real-time, for other objects tracking and counting (cars, taxi, bags, service personnel), for gender detection of human heads from any side on real video.
  • - Development real-time face recognition system for precise queue management
  • - Development predictive ML models for queues, waiting times, airport’s malls sales
  • - Analysis of the modern state of the problem, analysis patents and articles
Высшее образование
Национальный исследовательский университет «Московский институт электронной техники»
Факультет: Электроники и компьютерных технологий (ЭКТ)
Январь 1998—Январь 2001 (3 года)

Кандидат физ.мат. наук.