Senior Math Libraries Engineer, CPU and GPU Optimization

NVIDIA•Us, CA

17d

About The Position

NVIDIA is looking for an expert software engineer to help us deliver CUDA-X libraries across the NVIDIA CPU and GPU ecosystem. For over a decade, NVIDIA's accelerated computing platform has revolutionized HPC and AI with applications ranging from COVID-19 research to autonomous machines. Did you know that our team develops the GPU/CPU-accelerated mathematical libraries that make all of this possible? The hardware and software accelerated computing ecosystem is constantly evolving, including shifts towards hybrid backends, deep integration with high-level languages and ecosystems (such as Python, Numpy, JAX, MLIR…), and optimization at runtime for maximum flexibility and performance. Our libraries follow CUDA Everywhere approach to let developers use highly-optimized mathematical operations on all hardware available in NVIDIA ecosystem. You will be part of a team designing, developing, and optimizing math libraries for the future. If you are passionate about designing modern HPC libraries and want to build software that will stand the test-of-time as it accelerates countless applications, we might have the dream job you have been waiting for!

Requirements

PhD or MSc degree in Computer Science, Applied Math, or a related science or engineering field is preferred (or equivalent experience).
12+ years of experience designing and developing software for high-performance computing and/or AI applications.
Advanced C++ skills, including modern design paradigms (e.g., template meta-programming, RAII).
Parallel programming experience with CUDA, OpenCL or vector programming on CPU (AVX, NEON or similar).
Strong collaboration, communication, and documentation habits.
Experience with ARM, RISC-V and/or x86_64 CPU architectures.

Nice To Haves

Strong background in numerical methods (e.g., FFT, numerical linear algebra).
Programming skills with Python, and modern automation setups for both building software (e.g. cmake) as well as testing it (e.g. CI/CD, sanitizers).
Experience with cross-compilation, setting up CPU/GPU/accelerator (cross-)compilation toolchains and bringing existing codes to new architectures.
Background with CCCL, OpenMP, OpenACC, multi-threading, MPI, PGAS.
Experience with scientific and deep learning libraries and frameworks such as PyTorch, JAX, MKL, MAGMA, PETSc, Kokkos, etc.

Responsibilities

Design modern, flexible, and easy to use APIs and kernels for math libraries and lead design reviews with all collaborators.
Work closely with internal (e.g., Engineering, Product Management) and external partners such as researchers to understand their use cases and requirements.
Work with internal and external customers to deliver timely math libraries releases.
Become a domain expert by continuously surveying current trends in software systems.