09/06/2022 updated
100 % available
Computer Vision & Machine Learning Team Lead
Moscow, Russian Federation
Russian Federation
Computer VisionDrilling OperationsMachine LearningRisk PreventionSatellite ImageryOil and GasCommercial Awareness
I passionate about solving business cases using computer vision, mostly in a real-time manner.
My last experience includes building a human factor risk prevention on drilling operations (oil & gas), identifying cows and analyzing their behavior on milk farms, and designing a sports video analytics system.
Regarding hackathons, I made an ML solution to determine areas prone to waterlogging with satellite imagery and helped to find solutions to predict crop rotation in France. I have a team with the perfect balance of science and business.
We are looking for an interesting project that can make the world a better place.
My last experience includes building a human factor risk prevention on drilling operations (oil & gas), identifying cows and analyzing their behavior on milk farms, and designing a sports video analytics system.
Regarding hackathons, I made an ML solution to determine areas prone to waterlogging with satellite imagery and helped to find solutions to predict crop rotation in France. I have a team with the perfect balance of science and business.
We are looking for an interesting project that can make the world a better place.
Experience
Professional experience0 - 5 years
Languages
EnglishFluentRussianNative speaker
Project history
A machine learning solution to determine areas of agricultural lands that are prone to waterlogging with the help of satellite imagery.
Dataset ESRI shapefile containing data of waterlogged regions is collected by manual soil examination on an area of 1500 km².
We built web application allowing to select an area for examination and observe visual computation results with a GeoJSON export option. The computational algorithm uses segmentation analysis on top of the U-Net architecture of satellite imagery tiles and returns the result as a binary mask. Sentinel via Google Cloud is used to get satellite images.
Dataset ESRI shapefile containing data of waterlogged regions is collected by manual soil examination on an area of 1500 km².
We built web application allowing to select an area for examination and observe visual computation results with a GeoJSON export option. The computational algorithm uses segmentation analysis on top of the U-Net architecture of satellite imagery tiles and returns the result as a binary mask. Sentinel via Google Cloud is used to get satellite images.
Product design & business app development for real-time game data startup with computer vision and AI technologies. It enables coaches to see valuable match info in real-time and points after the match to improve athletes.
Product design & development from scratch of Real-time CV solution for oil & gas well drilling.
Task #1 Real-time emergency prevention: Machine is trained to understand drilling operations using SSD. 
It makes possible to control the drill string hold's delay that prevents sticking. Sticking is a severe accident which happens in 2/3 cases.
Task #2 Human factor risk monitoring: Tracking, Stabilization, Feature Detection and Optical Flow algorithms help to control the right number of turns of drill pipe while screwing. It avoids accidents at thousands of meters of depth.
Task #1 Real-time emergency prevention: Machine is trained to understand drilling operations using SSD. 
It makes possible to control the drill string hold's delay that prevents sticking. Sticking is a severe accident which happens in 2/3 cases.
Task #2 Human factor risk monitoring: Tracking, Stabilization, Feature Detection and Optical Flow algorithms help to control the right number of turns of drill pipe while screwing. It avoids accidents at thousands of meters of depth.