Boutique systems integration for HPC & AI
Bringing science and craft to HPC and AI systems integration.
Fluid Numerics helps teams design, build, and operate high-performance systems: GPU clusters, Slurm stacks, and reproducible ML pipelines. We work at the level of detail that keeps them reliable.
Cloud • Bare metal • Containers • ROCm & CUDA • Slurm
Ecosystem
Compute providers we work with
Tools we build with
Technologies we integrate
Languages we speak
Capabilities
Thoughtful systems work. Built by practitioners who care about reliability, performance, and the long tail of operations.
Cluster bring-up, done right
Bare metal or cloud, from pre-baked and tested images that ship with the scheduler and monitoring already in place. Every build is repeatable.
Researcher support
Open communication, office hours, and tutorials built around your workflows and your research software.
Slurm & workflow plumbing
Partitions, accounting, QoS, preemption, and job ergonomics, set up so that submitting a job is the easy part.
Reproducible software stacks
Versioned containers and modules, curated base images, and CI that proves the stack works before users do.
Performance as a practice
Profiling, tuning, and scaling guidance from microbenchmarks, mini-apps, and your own applications. The method and the measurements are written down.
Flexible management
Co-managed or fully managed engagements that fill the gaps in cluster operations, administration, and user support.
Tune in to our world
WFLD-FM, independent radio from Hickory, North Carolina.
Yes, we do that too.
A team with backgrounds spanning Applied Mathematics, Geophysical Fluid Dynamics, Scientific Computing, Computational Physics, Electronics, and Music is bound to have side projects.
Let’s build something solid.
Tell us what you’re running, or what you’re trying to run. We’ll help you design a system you can maintain and measure, and that will still be running in five years.