My basic training was in physics and applied mathematics (mainly statistical physics, hydrodynamics and turbulence). However, in the course of the last 17 years, since I moved to LANL, it gradually shifted towards the Science of Complex Engineered Systems. Today I spend 50% of my time on Energy Systems, 30% on Machine Learning, Optimization and Control Theory, devoting the remaining 20% to catching up on whatever is hot and interesting in Statistical Physics.
Some of the physics subjects which were dear to my heart ten years ago, like turbulence theory and fiber optics, remain on the back-burner mainly because my other engineering stories seem more exciting and timely.
All in all, I love to be in between different disciplines, as it gives me the opportunity to study, learn and challenge myself.
The energy system project has started some 10 years ago -- first as an attempt to find an application area for theoretical work on algorithms of discrete optimization/statistical inference. The very pragmatic initial interest has developed into a much broader endeavor, which matured through LANL LDRD project on Optimization and Control Theory for Smart (power) Grids and evolved into the ``Science of the Grid" project funded by DOE/OE. The main novel ingredient of this project consists in understanding of (and consequently optimizing and controlling) collective and stochastic processes in electric power systems and more generally energy systems (heating and gas transmission and delivery included). The project utilizes methods from applied mathematics and statistical physics normally used to study and optimize natural systems.The energy system project also required significant immersion into Stochastic Control and various related subjects in optimization and machine learning.
To explain some of my other interests, it may have a sense to briefly recall other steps in my carrier. I began analyzing dynamics of phase transitions and quantum spin systems. This naturally evolved into an in-depth study of turbulence, weak and strong, passive and active. Modern name of this field is Statistical Hydrodynamics, which also fits as a sub-field into Non-equilibrium Statistical Physics dealing with emergence and statistics of currents and topology and mixing in phase spaces. My first applied project examined effects of noise and disorder in Fiber Optics. I studied the ``fiber communication channel" and was surprised to learn that the error-correction can be efficient even without detailed statistical information about the channel known. This first encounter with error-correction developed into my second most active project (as of today) -- Physics of/and Algorithms, which focused on analysis of problems in machine learning (inverse and direct problems stated as graphical models), combinatorial and continuous optimization and control theory blending into this mix modern methods of statistical physics.