High-Performance GPU Kernel & CUDA Engineer (C++) Equity

SpaceX Super Intelligence; Seattle (WA); On-site; USD 180,000 - 440,000; Full time

Benefits offered by this job

Equity

Medical coverage

Vision coverage

Dental coverage

401(k) plan

Disability insurance

Life insurance

Discounts & Perks

Job summary

SpaceXSI is building one of the world’s largest AI supercomputers. You will own the raw GPU cluster and the platform layer, spanning from GPU kernel internals to large-scale orchestration and virtualization. The role aims to accelerate Grok's training speed and AI progress across the stack.

You will work across GPU kernels, memory hierarchies, and distributed systems to ensure fast, reliable inference and training at scale, collaborating with AI researchers to deliver scalable solutions.

Qualifications

• Deep low-level systems programming with C/C++/PTX/SASS.
• Experience with large-scale GPU clusters or distributed infra.
• Hands-on GPU kernel optimization with CUTLASS, Nsight.
• Track record building or running high-performance AI infra.

Responsibilities

• Design, build, and optimize massive GPU clusters for extreme-scale training and inference workloads.
• Develop and tune low-level CUDA kernels (GeMM, Attention) using CUTLASS, Tensor Cores, Nsight for max performance.
• Profile, debug, and eliminate bottlenecks across GPU memory, networking, filesystems, and multi-GPU operation.
• Collaborate with AI research teams to deliver production-grade performance and scalability.

Skills

Deep low-level systems programming

Large-scale GPU clusters

GPU kernel optimization

High-performance AI infrastructure

First-principles optimization

Tools

CUTLASS

Nsight

CUDA

PTX/SASS

Job description

SpaceXSI is building one of the world’s largest AI supercomputers. You will own the raw GPU cluster and the platform layer, spanning from GPU kernel internals to large-scale orchestration and virtualization. The role aims to accelerate Grok's training speed and AI progress across the stack.

You will work across GPU kernels, memory hierarchies, and distributed systems to ensure fast, reliable inference and training at scale, collaborating with AI researchers to deliver scalable solutions.