GPU Kernel Library Software Developer

Project Description:

Luxoft is looking for an individual to join an ambitious team developing Deep Learning and High-Performance Computing GPU kernels on the AMD Radeon Open Compute (ROCm) platform for MIOpen and Composable Kernel, AMD's Deep Learning primitives libraries which provides highly optimized implementations of different operators.

• https://github.com/ROCmSoftwarePlatform

• https://github.com/ROCmSoftwarePlatform/composable_kernel

• https://github.com/ROCmSoftwarePlatform/MIOpen

The successful person will be an experienced GPU-compute programmer with an eye towards hardware-aware performance optimizations.

Responsibilities:

The ideal candidate will be responsible for writing high performance GPU kernels for Machine Learning and Deep Learning Library: MIOpen and Composable Kernel

• Porting and optimizing algorithms for new GPU hardware

• Performing code reviews, building unit tests, authoring detailed documentation related to their work, and working with on-site and off-shore teams to deliver the software solutions on schedule.

• Playing a key role in all phases of the software development including system requirements analysis, coordinating feature design and development across functional and organization boundaries.

Mandatory Skills Description:

• Strong programming skills in modern C++ (templates, compile-time optimizations)

• In-depth knowledge about one of the parallel programming technologies CUDA/HIP/OpenCL/SYCL etc.

• Experience in parallel computing on multi-core/multi-node architectures (GPU/DL accelerators, computer clusters)

• Experience with parallel programming techniques and optimizations

• Understanding of Linear Algebra routines on tensors ("general algorithms" knowledge, not necessarily Linear Algebra)

• Good teamwork and interpersonal skills required

• Ability to work independently and within complementary teams

Nice-to-Have Skills:

• Demonstrate flexibility, strong motivation and a proven track record of meeting results-oriented deadlines.

• Detailed knowledge about GPU/accelerators hardware architecture from computational perspective

• Familiarity with deep neural network machine learning technologies, architectures and modern machine learning programming frameworks

• Experience working with and developing virtualization containers and package managers for code deployment

• Experience working with CPU/GPU assembly

• Basic understanding of Linux OS

• Basic knowledge of software development lifecycle, SW practices including debug, test, revision control, documentation, and bug tracking

• Experience using version control software such as Git

Languages:

English: C1 Advanced