Senior HPC CUDA Engineer for Geoscience & Seismic Imaging

EPAM; United States; Hybrid; USD 150,000 - 210,000; Full time

Benefits offered by this job

Relocation opportunities

International assignments

Corporate and social events

Unlimited access to internal learning

Job summary

EPAM is seeking a Senior HPC / CUDA Engineer to optimize GPU-accelerated scientific computing applications in geoscience, focusing on seismic imaging and Full Waveform Inversion. You will work on CUDA kernel optimization, C++/Python development, and distributed multi-GPU computing to boost performance and scalability.

The role requires deep CUDA knowledge, experience with MPI/NCCL/NVSHMEM, and a strong background in high-performance computing for large-scale workloads.

Qualifications

• Strong HPC and GPU-accelerated software development experience.
• Advanced CUDA programming and GPU optimization knowledge.
• Deep understanding of NVIDIA GPU architecture and memory hierarchy.
• Experience with GPU profiling and identifying bottlenecks.

Responsibilities

• Design, develop, and optimize CUDA kernels and GPU-accelerated apps.
• Analyze and improve performance across NVIDIA GPUs.
• Profile performance at kernel, PTX, and multi-GPU levels.
• Develop distributed multi-GPU solutions using MPI/NCCL/NVSHMEM.
• Improve seismic imaging and FWI workloads for scalability.
• Collaborate with domain experts to deliver technical solutions.

Skills

CUDA programming

GPU optimization

C++ development

Python

Performance profiling

Tools

MPI

NCCL

NVSHMEM

HDF5

Job description

EPAM is seeking a Senior HPC / CUDA Engineer to optimize GPU-accelerated scientific computing applications in geoscience, focusing on seismic imaging and Full Waveform Inversion. You will work on CUDA kernel optimization, C++/Python development, and distributed multi-GPU computing to boost performance and scalability.

The role requires deep CUDA knowledge, experience with MPI/NCCL/NVSHMEM, and a strong background in high-performance computing for large-scale workloads.