One of the best qualities that many of Akvelon’s employees share is their passion for technology that drives them to continue pushing themselves further in their field. One such employee is Anton Nesterenko, a senior data scientist at Akvelon, who is always taking on new challenges in machine learning. Anton has just won a top score in the Rosneft Seismic Challenge, a machine learning competition.
Rosneft Seismic Data Analysis Challenge
The Anton’s team had some steep competition, with 97 teams competing in the first round and 42 teams competing in the second round. One particularly impressive team that Anton faced was ZFTurbo, which was ranked top 8th in the world in Kaggle competitions.
Competitors were faced with a complex problem that involved using machine learning to estimate seismic horizons in a cube of amplitudes.
They were challenged with developing an algorithm to isolate seismic horizons in the cube of amplitudes, to solve the problem of data segmentation by image recognition.
Anton’s team, “IMDxD“, ranked in the top 30 teams in the 1st round, qualifying to compete in the second round, where they earned a rank of 10th place. Team IMDxD utilized machine learning in their solution to help with seismic cube segmentation, using a 2 model UNet and FPN ensemble.
When asked what the most challenging part of the competition was, Anton said “It was a hard competition because we had a very small dataset for training”. It was also Anton’s first experience working with segmentation, and he proved that he is a very quick learner.
Dedication to Machine Learning
Anton has proven to be very dedicated to the field of machine learning. This past year, he taught a 3-month machine learning course at Akvelon’s Ivanovo office, teaching over 50 Akvelon about several topics in machine learning including teaching with and without a teacher, intelligence data analysis, and working with a times series.
Anton also delivered a presentation on his solution at the Seismic Data Analysis Championship conference on Saturday December 21st, where he shared his research findings and algorithm with over 50 attendees.
We are very thankful to have such a driven individual on our team, and are very excited to see the cool machine learning projects and challenges that Anton works on next!
