GPU Performance Scientist: Transformers & CUDA Kernels

Luma; Greater London; Presencial; GBP 120.000 - 180.000; Jornada completa

Descripción de la vacante

Luma is seeking a senior performance engineer to accelerate its multimodal models. You will profile and optimize GPU, CPU, and accelerator code to train faster and deploy at scale without compromising quality.

You will write kernels and operations that extract maximum throughput from modern hardware. This is deep performance work requiring expertise in CUDA/Triton, transformer internals, and distributed multi-node deployment.

Formación

• Expert-level Triton/CUDA programming and GPU optimization.
• Strong PyTorch skills, including kernel development and custom operations.
• Proficiency with profiling tools (NVIDIA Nsight, torch profiler, custom tooling).
• Deep understanding of transformer architectures and attention mechanisms.

Responsabilidades

• Profile and optimize GPU/CPU/accelerator code for maximum utilization and minimal latency.
• Write high-performance PyTorch, Triton, and CUDA, dropping to custom operations when needed.
• Develop fused kernels and leverage tensor cores and modern hardware features across platforms.
• Optimize model architectures and implementations for distributed multi-node production deployment.
• Build performance monitoring and analysis tools and automation.
• Research and implement cutting-edge optimization techniques for transformer models.

Conocimientos

Triton/CUDA

GPU optimization

PyTorch kernel development

Profiling tools

Transformer architectures

Herramientas

Nsight

Torch Profiler

TensorRT

Descripción del empleo

Luma is seeking a senior performance engineer to accelerate its multimodal models. You will profile and optimize GPU, CPU, and accelerator code to train faster and deploy at scale without compromising quality.

You will write kernels and operations that extract maximum throughput from modern hardware. This is deep performance work requiring expertise in CUDA/Triton, transformer internals, and distributed multi-node deployment.