VK is one of the biggest social internet companies in Russia and Europe, it accounts for >90% of the Russian internet audience having revenue of $1.4 billion, it's in the top-50 most visited websites worldwide and top-10 in Russia in Russia with 71 million monthly active users. In my role within the Feed and Recommendations Team, I have gained extensive experience in machine learning product development, encompassing various stages:• Collecting datasets from scratch, including organizing the work of human annotators and utilizing crowdsourcing methods.• Building ETL pipelines using PySpark and Luigi to efficiently collect data from production.• Training machine learning models from scratch, primarily utilizing PyTorch and XGBoost.• Integrating data collection and model usage into PHP backend.• Deploying models into production using AB-tests and implementing monitoring solutions.Main projects:• Developing a new model for filtering clickbait posts, contributing to a better user experience.• Building a recommendation system aimed at promoting and supporting emerging authors.• Working on a model to understand and predict user thematic interests.• Currently, I am focused on developing a model for assessing content quality and enhancing the discovery of high-quality content.

During my internship in ITMO University laboratory, I had the opportunity to collaborate with VK, the largest social network in Eastern Europe. My main focus was on the topic of Active Learning and data labeling optimization. The objective of the research was to identify the most effective methods for posts thematic labeling.Key contributions during the internship include:- Developed a post thematic classification model using PyTorch.- Implemented various data labeling strategies and created a framework for their comparison.- Successfully identified the best solution that resulted in a 15% reduction in labeling efforts.This internship provided me with valuable experience in machine learning and data analysis, as well as the opportunity to work on a real-world problem in partnership with a renowned company like VK.