A small group, deliberately

Karthik works across both halves of what this lab does: deriving compressible flow models from kinetic theory, and engineering them to run efficiently on real hardware. It is an unusual combination to find in one person, and it is the reason the numerics here are shaped by what hardware is actually good at, while the implementation is shaped by what the physics actually needs.

His doctoral research at the University of Geneva developed compressible lattice-Boltzmann methods for GPU execution — positivity-preserving formulations for high-speed flow, and the moment-space non-equilibrium indicator that the lab’s face-local shock sensor descends from. He trained first as an aerospace engineer in propulsion, with research on lean hydrogen combustion and on plasma–fluid interaction at Institut P’.

He taught parallel programming and probabilistic algorithms at Geneva for three years. That is where the lab’s insistence on a single source for CPU and accelerator comes from — it is far easier to defend a result when there is only one implementation to defend.