Experience

Celsus

• Russian Federation
• Information Technology & Services
• 1 - 100 Employee

• Machine Learning Engineer (Medical imaging)

  • Jan 2022 - Present

  I work in a team developing systems for automatic cancer detection in mammogram images. This is the best AI system for mammography on the market with hundreds of thousands of real studies processed.• Designed and implemented from scratch an image annotation storage system on PostgreSQL. Moved all the stages of the training pipelines to it. Developed a dashboard on Dash to display annotations from the database• Researched and tested hypotheses to improve the accuracy of models and their architecture for object detection, segmentation and classification tasks• Transformed a heavily customized object detection architecture based on Faster R-CNN into instance segmentation Show less

• Machine Learning Engineer (Video analytics)

  • May 2021 - Jan 2022

  I was part of the team developing a video analytics system for intensive human service sites.• Trained detection, classification and re-identification models• Implemented active learning pipeline for object detection (EfficientDet) based on GMM• Improved tracking algorithms (DeepSORT, ByteTrack)• Initiated and managed the migration of the inference pipeline to the NVIDIA Triton Inference Server. As a result of the migration, processing speed has more than doubled• Organized storage of the results of the video analytics algorithms in PostgreSQL and developed dashboards on Dash to visualize them Show less



Far Eastern Federal University

• Public Relations and Communications Services
• 100 - 200 Employee

• Robotics Software Engineer

  • Jan 2017 - Sep 2020

  I worked in an industrial robotics laboratory that focuses on the development of robotic cells for an aircraft factory. I developed software for laser scanning and positioning of metal parts for machining with KUKA robots. • Developed an algorithm to synchronize robot and sensor measurements • Designed and implemented an algorithm to calibrate the precise position of the scanner on the robot • Developed algorithms for point cloud registration between 3D measured points and CAD models of fabricated parts • Created a GUI to visualize measured point clouds, robot, the results of calibration and model registration algorithms • Implemented several other MVPs: online adjustment of the robot's machined trajectory based on scanning and hand guiding control of the robot Show less