HPC Performance Engineer - AI Training & GPU Optimization

CuspAI in Cambridge/London/Amsterdam/Berlin is seeking a Performance Engineer to bridge frontier AI models with our high performance compute. You will profile training runs, identify bottlenecks, and design optimizations to scale large models for chemistry and physics.

You will collaborate with AI researchers and the Infrastructure team to optimize CUDA kernels, NCCL ops, and data pipelines, plus build real time dashboards to maximize MFU and accelerator utilization.