GPU Performance Engineer for Transformer Models

Openkrill; United States; Teletrabalho; USD 180 000 - 240 000; Tempo integral

Resumo da oferta

Luma is seeking a performance-focused engineer to accelerate multimodal models across GPU, CPU, and accelerators, writing kernels and optimized ops.

You will work with Triton, CUDA, PyTorch, and distribution across multi-node deployments, delivering measurable latency and throughput gains.

If you thrive on deep transformer internals and profiling, this role offers a chance to shape production-scale AI systems.

Qualificações

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

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 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.

Conhecimentos

Triton/CUDA

PyTorch kernels

GPU optimization

Profiling tools

Transformer attention

Ferramentas

NVIDIA Nsight

Torch profiler

Custom tooling

Descrição da oferta de emprego

Luma is seeking a performance-focused engineer to accelerate multimodal models across GPU, CPU, and accelerators, writing kernels and optimized ops.

You will work with Triton, CUDA, PyTorch, and distribution across multi-node deployments, delivering measurable latency and throughput gains.

If you thrive on deep transformer internals and profiling, this role offers a chance to shape production-scale AI systems.