Research Scientist / Engineer – Performance Optimization

Luma

Remote, EU

Remote

Who this role is best for

Candidates with deep GPU-optimization expertise and transformer knowledge will find this performance-focused role in a remote EU-based company.

Best fit for

Individuals with advanced CUDA and Triton experience and a history of impactful kernel development
— “Write high-performance PyTorch, Triton, and CUDA, dropping to custom operations when needed.”
Candidates who have optimized transformer models and understand attention mechanisms deeply
— “Deep understanding of transformer architectures and attention mechanisms.”
Professionals skilled in profiling and performance analysis with experience in modern GPU tools
— “Proficiency with profiling tools (NVIDIA Nsight, torch profiler, custom tooling).”

Things to consider

The role demands a high level of technical depth in GPU and distributed systems
— “This is deep performance work: fused kernels, tensor cores, Triton and CUDA, distributed multi-node deployment.”

How to stand out

Highlight end-to-end optimization projects with measurable latency or utilization improvements
— “Land a kernel or architecture optimization that measurably improves utilization or latency.”
Showcase experience with building monitoring tools and automation for performance tracking
— “Build performance monitoring and analysis tools and automation.”
Demonstrate familiarity with leveraging tensor cores and fused operations across platforms
— “Develop fused kernels and leverage tensor cores and modern hardware features across platforms.”

Pace · Fast PacedCollaboration · HighAutonomy · HighDecision Impact · Company

Derived from job-description analysis by Serendipath's career intelligence engine.

What success looks like

Profile and optimize GPU/CPU/accelerator code
Write high-performance PyTorch, Triton, and CUDA
Develop fused kernels and leverage tensor cores
Optimize model architectures and implementations for distributed multi-node production deployment

Typical background

Expert-level Triton/CUDA programming and GPU optimizationStrong PyTorch skills, including kernel development and custom operationsProficiency with profiling tools (NVIDIA Nsight, torch profiler, custom tooling)

Skills & requirements

Required

GPU OptimizationCUDA ProgrammingPyTorchProfiling ToolsTransformer Architectures

Preferred

Compilers And ExportersInference Workload OptimizationTriton Compiler And Kernel Fusion TechniquesWarp-level Intrinsics And Advanced CUDA Optimization

Stack & domain

GPU OptimizationCUDAPyTorchTritonTensor CoresDistributed Multi-node DeploymentProfiling ToolsNvidia NsightTorch ProfilerCustom OperationsFused KernelsTransformer ArchitecturesAttention MechanismsCompilersExportersInference WorkloadsLatencyThroughputWarp-level IntrinsicsAdvanced CUDA OptimizationMultimodal ModelsAIMachine Learning

About the role

Original posting from Luma via Ashby

You'll make Luma's multimodal models fast — profiling and optimizing GPU, CPU, and accelerator code so they train efficiently and deploy at scale without sacrificing quality. You'll write the kernels and operations that get the most out of the hardware.
This is deep performance work: fused kernels, tensor cores, Triton and CUDA, distributed multi-node deployment. It fits someone with expert GPU-optimization skills and a deep understanding of transformer internals. If you're not at home in CUDA, Triton, and profilers, this is the wrong depth.
What You'll Own
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.
First 90 Days
One way the first 90 could unfold.
Days 1–30 — Immerse & Diagnose: Profile the current training and inference paths and find the biggest performance wins.
Days 30–60 — Ship & Validate: Land a kernel or architecture optimization that measurably improves utilization or latency.
Days 60–90 — Scale & Systemize: Build the monitoring and automation that keeps performance gains from regressing.
What You Bring
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.
Nice to Have
Experience with compilers and exporters (torch.compile, TensorRT, ONNX, XLA).
Experience optimizing inference workloads for latency and throughput.
Triton compiler and kernel fusion techniques.
Knowledge of warp-level intrinsics and advanced CUDA optimization.
About Luma: Luma's mission is to build unified general intelligence that can generate, understand, and operate in the physical world. We believe multimodality is critical for intelligence — the next step beyond language models comes from vision. Luma is an equal opportunity employer.

Source: Luma careers (Ashby)