Senior GPU Kernel Developer

Luxoft; Torino; In loco; EUR 50.000 - 70.000; Tempo pieno

Descrizione del lavoro

A leading tech company in Italy seeks an experienced developer with GPU compute skills to lead the optimization of HIP kernels on AMD GPUs. This role involves close collaboration with development teams, debugging, and enhancing GPU-accelerated applications. The ideal candidate should have strong proficiency in C++ and CUDA/HIP, with a solid understanding of GPU architectures and optimization techniques. Familiarity with AI/ML techniques is also preferred.

Competenze

• Experience in GPU compute and performance profiling.
• Strong background in GPGPU applications.
• Familiarity with optimization techniques.

Mansioni

• Optimize HIP kernels for AMD hardware.
• Collaborate with teams on GPU-accelerated applications.
• Debug, profile, and fine-tune code for performance.

Conoscenze

C++ proficiency (C++ 17)

CUDA or HIP programming

GPU architectures understanding

Parallel programming models

Problem-solving skills

AI/ML/DL/NN/NLP/Computer Vision experience

Python

Strumenti

Linux

gdb/LLDB

Jinja2 or similar templating engines

Descrizione del lavoro

Project description

Luxoft is searching for talented developers with GPU compute and performance profiling experience to join the rapidly growing team. We are seeking an experienced individual proficient in GPGPU applications to join our team. The primary responsibility of this role will be to lead the effort in optimizing HIP kernels on AMD GPUs. The candidate should possess a strong background in GPU computing, parallel programming, and a deep understanding of CUDA or HIP frameworks. Additionally, familiarity with optimization techniques is highly desirable.

Responsibilities

• The main task will be to help optimize HIP kernels for specific AMD hardware. Collaborate with development teams to optimize and enhance GPU-accelerated applications. Debug, profile, and fine-tune code for performance improvements. Stay updated with the latest advancements in GPU architectures and programming models.

Skills

Must have

• Proficiency with C++ and low-level programming (at least C++ 17). Proficiency in CUDA or HIP / ROCm programming. Solid understanding of GPU architectures, parallel programming models, and optimization techniques. Strong problem-solving skills and the ability to work in a collaborative environment. Experience with AI/ML/DL/NN/NLP/Computer Vision. Python.

Nice to have

• Linux. CPU Intrinsics (AVX/SSE). GPU Assembler. Profiling. gdb/LLDB. Jinja2 or similar templating engines.