Senior GPU Linear Algebra Kernel Engineer

NVIDIA; California (MO); On-site; USD 184,000 - 357,000; Full time

Job summary

NVIDIA seeks software engineers to advance dense linear algebra kernels for high-performance libraries such as cuSOLVER. You will design, implement, and optimize GPU kernels for triangular factorizations, eigenvalue computations and SVDs, contributing to state-of-the-art libraries.

You will work with product management and partners to meet performance requirements and drive technical roadmaps while improving quality and maintainability of the libraries.

Qualifications

• PhD or MSc degree in Computational Science and Engineering, CS, Applied Mathematics, or related field (or equivalent).
• 5+ years of experience in high-performance numerical linear algebra software using C++ and parallel programming (CUDA, MPI, OpenMP, OpenACC, pthreads).
• Strong fundamentals in numerical methods and linear algebra solvers, eigen/singular value decompositions.
• Experience with dense linear algebra libraries such as BLAS, LAPACK and parallel counterparts.

Responsibilities

• Design, implement and optimize scalable high-performance numerical dense linear algebra software on GPUs.
• Provide technical leadership to library engineers, QA engineers, and interns.
• Collaborate with product management and partners to meet feature and performance requirements.
• Improve library quality, performance and maintainability through re-architecting and innovative practices.

Skills

C++

CUDA

MPI

OpenMP

OpenACC

Pthreads

Numerical linear algebra

BLAS/LAPACK

Education

PhD or MSc in Computational Science & Engineering / CS / Applied Math (or equivalent)

Tools

BLAS

LAPACK

PBLAS

SCALAPACK

CUDA toolkit

Job description

NVIDIA seeks software engineers to advance dense linear algebra kernels for high-performance libraries such as cuSOLVER. You will design, implement, and optimize GPU kernels for triangular factorizations, eigenvalue computations and SVDs, contributing to state-of-the-art libraries.

You will work with product management and partners to meet performance requirements and drive technical roadmaps while improving quality and maintainability of the libraries.