Portable GPU Programming

Paswey Supercomputing Centre / CSIRO

Fluid Numerics began working with CSIRO and the Pawsey Supercomputing Centre in Australia in 2022. Our role is to provide researchers with highly focused mentored sprints that accelerate the transition to AMD's ROCm ecosystem to ensure success on Setonix.

Advanced Micro Devices (AMD)

Fluid Numerics has maintained a working relationship with AMD since 2018. During this time, we have provided consulting and training services in High Performance Computing. In 2022, we started working with AMD on ROCm training resources, tutorials, and a ROCm developer certification program.


Google Cloud

Fluid Numerics has been a Google Cloud and Google Workspace partner since 2019. We offer solutions on Google Cloud's Marketplace and have provided professional services Google Cloud to assist in the development and benchmarking of solutions deployed with Slurm-GCP.


Fluid Numerics began working with Midjourney in 2022. We design, deploy, and manage a cloud-native HPC cluster that provides access to more than 30 machine topologies on Google Cloud and integrates with the Weka File System. Additionally, we provide user support services to help Midjourney's creative team best leverage this solution.

Academic Research

Florida State University

Fluid Numerics has been working with Florida State University since we opened shop in 2018. In this role, we have participated with university researchers, Dr. Bill Dewar and Dr. Nicolas Wienders, on studying the dynamics of the Gulf Stream separation funded by the National Science Foundation. In addition to generating 300m resolution numerical simulations of the Gulf Stream, we have been developing a mathematical framework for diagnosing spectra in irregular geometries.

Los Alamos National Laboratory

Fluid Numerics has been working with Los Alamos National Laboratory since 2018. We have been working with Dr. Wilbert Weijer on developing capabilities to simulate passive tracers using graph theory to diagnose advection-diffusion operators from state of the art climate simulation models. We continue to seek avenues to apply this methodology to continue work on the FEOTS toolkit.