• Utilized reinforcement learning, evolutionary strategies, deep learning, and classical ML to develop an automated quantitative trading system based on time series analysis• Wrote the whole pipeline from data downloading and preprocessing to inference and analyzing the outputs• Achieved huge improvements in quality of the predictions on the backtest• Mentored junior engineers• Conducted ML interviews for middle and senior level positions

• Created several algorithms for estimating the precision of quantum tomography• Published several scientific papers based on conducted work in Q1 journals• Wrote a framework for quantum computations and tomography from scratch on Python

• Completely rewrote the deep learning computer vision pipeline and cloud job submission infrastructure• Designed custom neural network architectures for specific problems• Achieved large improvements in classification performance and embedding quality• Applied techniques such as self-supervised representation learning, adversarial learning, recalibration, and aleatoric and epistemic uncertainty

The goal of the project was to analyze public information (subscriptions and likes) of people's profiles on social networks to predict their preferences in different topics.• Developed algorithms of data extraction from social networks• Used neural networks and classical ML methods to predict users' preferences from the extracted data