George Nigmatulin

Project Title: Generative AI for financial data: From Price-Impact Consistency to Reinforcement Learning Driven Trading

I specialise in machine learning for financial data. My work combines generative AI, Limit-Order-Book microstructure, and Reinforcement Learning: I build synthetic order-flow generators that reproduce market impact and other stylised facts, then use them as controlled RL simulators to learn policies that transfer to live trading. This helps pension funds, asset managers, and regulators cut crash risk and uphold financial stability. My broader interests include machine learning mid/low-frequency strategies spanning causal inference, regime modelling and portfolio optimisation

Before starting my DPhil, I earned an MSc in Applied Mathematics and Physics from the Moscow Institute of Physics and Technology (MIPT, 2022). My core industry experience is Machine Learning Engineer at Samsung R&D, where I delivered full-stack smartwatch ML -data engineering, time-series modelling, and on-device deployment. In academic research, I've been engaged in the fields of Uncertainty Quantification, Anomaly Detection, and Deep Learning in the Laboratory of Applied Research for Structured Data Statistics at Skoltech. I've also worked on predictive modeling for a Top-1 regional bank, predicting the cash requirements for ATMs using time-series forecasting algorithms and creating a system to optimize cash management.