Plasmion AI was founded on the conviction that nuclear fusion's ultimate obstacle is not magnetic hardware, but real-time computational prediction and control.
For seventy years, the dream of unlimited, carbon-free energy from magnetic fusion has been stalled by plasma turbulence. As magnets grew stronger, the plasma grew exponentially more unstable. Traditional supercomputing could simulate plasma after the fact—over weeks of cluster time—but proved completely useless inside the reactor’s millisecond control loops.
In 2024, our founders realized that continuous Fourier Neural Operators (FNO) trained on millions of multi-scale synthetic gyrokinetic discharges could solve nonlinear magnetohydrodynamic PDEs 100,000x faster, enabling sub-microsecond closed-loop inference.
Today, Plasmion AI provides the real-time neural operating system that allows private and public fusion reactors to operate safely at burning plasma pressures ($Q > 10$).
Physics-Constrained AI: We do not deploy black-box models. Every loss surface strictly incorporates $\nabla \cdot \mathbf{B} = 0$ and conservation of magnetic helicity.
Deterministic Execution: Operating at 200 kHz demands nanosecond hardware predictability with zero kernel jitter.
Multi-Device Generality: Our foundation models transfer zero-shot across diverse tokamak geometries.
Led by senior computational plasma physicists and extreme-scale distributed systems architects.
Founder & Chief Executive Officer
Ph.D. in Computational Plasma Physics & Applied Mathematics (MIT / Princeton Plasma Physics Laboratory). Former Principal Scientist at Max Planck Institute for Plasma Physics. Lead developer of spectral PDE surrogates.
Chief Science Officer & Co-Founder
Ph.D. Imperial College London. 12+ years at the ITER Theory & Modelling Division specializing in macroscopic tearing modes, non-inductive current drive, and resonant magnetic perturbation coils.
VP of AI Systems & HPC
M.S. Stanford University. Expert in distributed tensor accelerator clusters, custom high-throughput DMA interconnects, and low-latency microkernel development for FPGA and hardware accelerators.
Head of Commercial & Strategy
MBA Harvard Business School, B.S. Nuclear Engineering. Former director of strategic partnerships in advanced nuclear energy infrastructure and public-private utility ventures.
Chair of Theoretical Plasma Physics, EPFL
Pioneer in gyrokinetic simulation codes and drift wave turbulence theory. Adviser on Sobolev loss regularizers.
Senior Fellow, Tokamak Control Consortium
30+ years overseeing magnetic diagnostic integration and high-temperature superconducting coil protection.
Emeritus Lead, Superconducting Magnet Laboratory
Expert in high-field REBCO tape magnet quench dynamics and fast inductive power supply switching.