HPC GPU Engineer

Contract HPC Performance Engineer - Cluster Optimization, GPU/CFD WorkloadsAn Enterprise client are hiring for a Contract HPC Performance ConsultantHourly rate $90-120 hourFully remoteProject; Diagnose and resolve indeterminate performance issues on a newly built HPC cluster, improving scaling, stability, and job scheduling across multiple engineering groups.Cluster Diagnostics – Identify root cause of inconsistent scaling when multiple jobs run concurrently across nodes – goal: eliminate unpredictable performance drops (e.g. single job stable, second job halving through put)Job Scheduling & Configuration – Optimize scheduling and cluster configuration across six compute nodes and two GPU nodes – goal: maximize utilization and stability for concurrent workloadsNetwork Fabric Tuning – Assess and tune RDMA over Converged Ethernet (no Infiniband present, Dell 100GbE switches) – goal: determine network fabric's role in performance bottlenecks and inform expansion/replacement decisionsSoftware-Specific Optimization – Tune cluster and job configs for each CFD/simulation package individually – goal: ensure each tool runs at optimal performance rather than "just working"Workflow & Script Support – Refine and maintain existing scripts and environment modules – goal: enable broader adoption across engineering groups beyond the two primary current usersTools/Environment:Nvidia Base Command Manager (formerly Bright Cluster Manager)AMD chipset hardwareStar CCM+, Fluent, Abacus, Converge, CFD GT Suite, JMag Electromagnetic, AVLVSMRDMA over Converged Ethernet, Dell 100GbE switchingSix compute nodes, two GPU nodesEngagement: Approx. 3–6 months, hands-on, iterative problem-solving alongside internal engineering teams (thermal, mechanical, internal combustion, vehicle simulation, electronics groups).Contract HPC Performance Engineer - Cluster Optimization, GPU/CFD Workloads