Hi, 👋 I’m Jinsu Kim. I’m a Ph.D student of Mechanical and Aerospace Engineering at Princeton University. I received my B.S.and M.S. in Nuclear Engineering and Physics at Seoul National University. My research focuses on plasma physics, reduced-order modeling, and optimal control.

I am particularly interested in the potential of machine learning for scientific and engineering applications. My previous work on data-driven modeling for nuclear fusion plasma includes: (1) Plasma instability modeling for disruption prediction, (2) PINN-based simulation for profile reconstruction, (3) Optimal control of plasma operation and instability suppression with RL, and (4) Reactor design optimization.

My research goal is to bridge the gap between physics and data-driven modeling. Currently, I am working on two main topics: (1) Symplectic model reduction for nonlinear Hamiltonian systems, and (2) Structure-preserving model reduction for Vlasov-Possion plasma systems. More broadly, my research interests include data assimilation for magnetohydrodynamic plasmas and optimal control of kinetic plasma systems for instability suppression.

My computational work is available on Github. If you are interested in my research, please feel free to explore my work or contact me—comments and discussions are always welcome.

Research area

Dynamics and Control

📌 Model order reduction on nonlinear dynamic systems

• Symplectic model order reduction for nonlinear Hamiltonian systems
  • Variants of Proper Symplectic Decomposition: Extension of Cotangnent lift
  • Integration with Stucture-preserving DEIM for reduction of nonlinearity computation
• Structure-preserving model reduction for Vlasov-Poisson plasma system
  • Development of Particle-In-Cell simulation in electrostatic plasma system
  • Symplectic model reduction for preserving Hamiltonian form in reduced space

📌 Computational plasma physics

• Symplectic integration of Particle-In-Cell method for plasma kinetic simulations
  • Development on PIC with symplectic integration in electromagnetic plasma system
  • Spectral solver for electromagnetic PIC and application of parallel computation

Nuclear Fusion

📌 Data-driven modeling for fusion plasma and optimized control

• Disruption prediction in KSTAR tokamak plasma with Deep Learning
  • Development of multi-modal deep neural network with multiple signals and IVIS for predicting thermal quench
  • Uncertainty modeling for high precision and causality estimation in disruption prediction with Bayesian deep learning
• Data-driven modeling and control for tokamak plasma operation
  • Development of physics-informed neural network for tokamak plasma simulation (Grad-Shafranov Physics-Informed Neural Network: GS-PINN)
  • Investigation for multi-objective plasma control with reinforcement learning

📌 Reactor Design Optimization

• Design optimization of a tokamak reactor with data-driven approaches
  • Development of tokamak reactor design computation code
  • Reactor design optimization for high performance and sustainable plasma operation based on bayesian optimization and reinforcement learning

AI in Nuclear Fusion: Bridging the gap between science and engineering