George Miloshevich

My research focuses on understanding and predicting complex processes in turbulent plasmas, particularly in space. I combine theory, numerical simulation and data-driven methods, with work ranging from modelling to statistical models.

I am an Associate Editor of JGR: Machine Learning and Computation and Project Manager of ASAP, a project developing autonomy for space missions. ASAP builds neural networks, along with the software and hardware needed to run them onboard, so that spacecraft can analyse and compress their observations in orbit rather than downlinking everything to Earth.

Research interests

• Turbulence and reconnection in space plasmas. Cascades and coherent structures in kinetic Alfvén-wave turbulence, and energy channels and pressure–strain interaction in magnetosheath turbulence. I study these using hybrid and implicit fully kinetic particle-in-cell (PIC) codes, extended magnetohydrodynamic (MHD) theory and gyrofluid simulations.
• Data-driven plasma modelling. Kinetic corrections to fluid models, including neural-network, analytic and symbolic-regression closures learned from fully kinetic simulations, and equations of state for reduced-order modelling.
• Space weather. Forecasting of coronal mass ejections (CMEs), solar flare classification and automated detection of coronal structures.
• Onboard algorithms for space missions. Neural networks running on field-programmable gate arrays (FPGAs) that detect solar structures and compress data, easing downlink bottlenecks on resource-limited spacecraft and delivering more science per bit of telemetry.
• Extreme events in environmental science. Forecasting long-lasting European heatwaves weeks ahead using convolutional neural networks and stochastic weather generators, and estimating the return times of unprecedented events with rare-event algorithms and extreme value statistics.
• Hamiltonian methods. Structure-preserving closures and reductions of extended MHD and Vlasov–Maxwell systems.

I received my PhD from the University of Texas at Austin, where I worked on Hamiltonian descriptions of microscopic kinetic effects in turbulent collisionless plasmas, combining analytical theory with numerical simulation. Before returning to plasma physics, I spent several years in environmental data science at ENS de Lyon and CEA Saclay, developing data-driven forecasts of extreme European heatwaves.

I come from a small country nestled in the Caucasus. Having grown up in a diverse family, I have always enjoyed and valued multicultural environments.

Find out more about me on my Personal Website