Summary

Fuse Energy is hiring a CUDA Engineer to develop and optimise low-level GPU code for transformer inference workloads. The role focuses on custom CUDA kernels, GPU performance tuning, quantisation, mixed precision, and high-performance compute infrastructure, with benefits varying by location.

Responsibilities

• Write and optimise custom cuda kernels
• Profile kernels and remove performance bottlenecks
• Apply kernel fusion across inference pipelines
• Optimise memory access and hierarchy usage
• Implement quantised mixed-precision inference kernels
• Build caching mechanisms for autoregressive decoding
• Benchmark kernels and write regression tests