Senior AI Performance Engineer - GPU & Triton Optimizations

lumalabs-ai; San Francisco (CA); On-site; USD 237,000 - 395,000; Full time

Job summary

Luma AI is hiring for a Performance Optimization role focused on making multimodal models faster and more scalable. You will work closely with research and engineering to optimize training and inference, write high-performance kernels, and push transformer models toward lower latency and higher throughput.

The role emphasizes deep CUDA/Triton expertise, PyTorch proficiency, and the ability to deploy efficient, production-grade optimizations across hardware platforms.

Qualifications

• Expert-level Triton/CUDA programming experience.
• Strong PyTorch skills and ability to develop custom operations.
• Experience with PyTorch kernel development.
• Proficiency with profiling tools (NVIDIA Nsight, torch profiler, custom tooling).
• Deep understanding of transformer architectures and attention mechanisms.
• Preferred: experience with torch.compile, TensorRT, ONNX, XLA.

Responsibilities

• Profile and optimize GPU/CPU/Accelerator code for maximum utilization and minimal latency.
• Write high-performance PyTorch, Triton, CUDA, deferring to custom PyTorch operations if necessary.
• Develop fused kernels and leverage tensor cores and modern hardware features for optimal hardware utilization on different hardware 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 model

Skills

Triton/CUDA programming

PyTorch

PyTorch kernel development

Profiling tools

Transformer architectures

Tools

NVIDIA Nsight

torch profiler

custom tooling

Job description

Luma AI is hiring for a Performance Optimization role focused on making multimodal models faster and more scalable. You will work closely with research and engineering to optimize training and inference, write high-performance kernels, and push transformer models toward lower latency and higher throughput.

The role emphasizes deep CUDA/Triton expertise, PyTorch proficiency, and the ability to deploy efficient, production-grade optimizations across hardware platforms.