Member of Technical Staff, Performance Engineering

The role

This person will own the speed of a workload end to end, from application code and data path through the scheduler to GPU kernels, across AI and HPC. The work is to make one cluster produce more physics, and each fix must be measured on real jobs.

What you will do

• Find and fix slow training loops, simulations or served models at the layer that causes the slowdown, and leave a benchmark behind.
• Specify, generate or write, verify and profile GPU kernels for physics campaigns and learned models.
• Go deep in one lane: multi-node AI training, inference serving at scale, or HPC simulation at scale.
• Measure the cluster on real jobs to inform queue policy and capacity decisions.

Required qualifications

• Three or more years of whole-stack performance work, including at least one layer that is not the GPU.
• Three or more years of systems programming in C++, CUDA or Triton, including reading and verifying code written by others.
• Experience with profilers such as Nsight Systems and Compute, torch profiler, perf, or NCCL tests.
• Scale experience in multi-node AI training, inference serving at scale, or HPC simulation at scale, with appetite for the others.
• Must be authorized to work in the United States.
• Must be a U.S. person if required by export control law, or eligible for any required U.S. government export authorization.

Nice to have

• Kernels written in CUDA C++ or Triton that other people's systems depend on.
• A second lane from the list of AI and HPC areas.
• Source-level work in OpenFOAM, LAMMPS, GROMACS, or comparable codes; PETSc, Trilinos, MAGMA or comparable libraries.
• Kueue or Slurm queue policy on a shared cluster.
• Experience turning research code into a maintained library that scientists depend on.

Benefits

• Salary
• Benefits
• Meaningful early-stage equity

Before you apply

• PSI is not able to sponsor or take over sponsorship of employment visas for this role, now or in the future.
• PSI may require the person in this role to be a U.S. person under export control laws, and will confirm eligibility with each candidate.

About the company

Physical Superintelligence is a startup building AI systems to discover new physics at scale. Its stated mission is to discover and commercialize transformative physics breakthroughs with artificial superintelligence, safely and for broad public benefit.