Artem Kurylev

Experience

• 

  Thomas More-hogeschool

  Feb 2018 - Apr 2018

  Intern Student

  Worked on web-based application dedicated to the needs of people with intellectual disabilities , aloowing them to have better conversations

• 

  EORA

  Jun 2020 - Dec 2020

  * Trained algorithms for digits classification to improve quality of digital and analogue water and electricity meter reading recognition * Warped deep learning models for production usage with TorchScript and ONNX* Trained and deployed model for background removal * Successfully implemented and trained algorithm for face landmark detection and achieved metrics and quality described in the article

  • Computer Vision Engineer

    Jul 2020 - Dec 2020

  • Computer Vision Engineer Intern

    Jun 2020 - Jun 2020
• 

  CLOFIT ME

  Oct 2020 - May 2022

  Computer Vision Engineer

  * Implemented several pipeline blocks for 2D virtual try-on, such as super resolution, landmark estimation, 2D human segmentation and clothes segmentation, trained and integrated model for 2d clothes landmark estimation. * Retrieve and preprocess data for 2d virtual try-on model training via scrapping shopping websites * Implemented several stages for auto-labelling pipeline to for continuously retraining model with Airflow * Created blending mechanism for improve quality of faceswap algorithm Show less

• 

  Roadly, Inc.

  May 2022 - now

  Machine Learning Engineer

  * Setting up and maintaining data-science workspace with several GPUs for experiments and development with Kubernetes, ClearML-Server, Traefik, Minio S3 * Created several pipelines with Airflow for car driving videos processing including data preprocessing, model inference, and storage downloads/uploads * Designed Postgres + Postgis database for storing geographical data with * Created several methods for fastapi application to retrieve processed data* Fine-tuned several models based on MMCV framework to improve quality of road distress segmentation* Trained multiple Pytorch models for raveling distress severity classification* Improved performance of video processing pipelines Show less