Remote AI Performance Engineer: High-Performance Kernels

Modular; United States; Remote; USD 180,000 - 270,000; Full time

Benefits offered by this job

RSU grants

Team onsites (Los Altos / other cities

Healthcare coverage

Job summary

Modular is seeking an AI Performance Engineer to join the ASIC Performance team. You will design, implement and tune high-performance attention kernels and a broader set of AI kernels for new accelerators.

You will translate algorithmic requirements into production-grade implementations for GPUs, NPUs, ASICs, and emerging accelerators. You will help shape Modular’s Attention framework, build reusable abstractions, and work with hardware partners to optimize end‑to‑end performance.

Qualifications

• 5+ years in HPC, AI kernel development, or accelerator programming.
• Experience writing production GPU/accelerator kernels with attention to memory movement.
• Knowledge of modern attention algorithms and their hardware tradeoffs.

Responsibilities

• Design and implement high-performance attention kernels for prefill and decode across variants.
• Build and optimize general AI kernels like matrix multiply, softmax, normalization, and embedding.
• Contribute to the Attention framework with hardware-specific optimization.
• Profile workloads, identify bottlenecks, and tune against hardware limits and baselines.
• Develop tests, benchmarks, and performance analysis for multiple model shapes.
• Collaborate with Modular teams to improve software stack and share learnings.
• Work with hardware partners to validate toolchains and interfaces.
• Write design docs, participate in reviews, and communicate performance insights.
• Mentor teammates and raise engineering bar for kernel quality.

Skills

GPU kernels

Performance profiling

Attention algorithms

Heterogeneous architectures

CUDA/Triton/SYCL/OpenCL

Low-level software

Mentoring

Education

Advanced degree in CS/EE/CE

Tools

CUDA

Triton

CuTe/CUTLASS

MLIR/LLVM

GPU acceleration toolchains

Job description

Modular is seeking an AI Performance Engineer to join the ASIC Performance team. You will design, implement and tune high-performance attention kernels and a broader set of AI kernels for new accelerators.

You will translate algorithmic requirements into production-grade implementations for GPUs, NPUs, ASICs, and emerging accelerators. You will help shape Modular’s Attention framework, build reusable abstractions, and work with hardware partners to optimize end‑to‑end performance.