Heterogeneous Computing Expert

Pinnacle Lead Consulting Limited; Hong Kong; On-site; HKD 900,000 - 1,100,000; Full time

Job summary

Pinnacle Lead Consulting Limited is seeking a Heterogeneous Computing Expert to advance AI platforms. You will collaborate with deep learning engineers to optimize software/hardware co-design and accelerate computer vision and speech models on CPU, GPU, DSP and embedded chips.

The role requires 5+ years of C/C++ development, hands-on heterogeneous programming (e.g., CUDA, OpenCL, OpenGL), familiarity with common DL models and platforms, and experience with compilers like TVM, Glow and LLVM.

Qualifications

• Solid computer foundation with knowledge of computer architecture and parallel computing basics.
• 5+ years of C/C++ development experience, proficient in at least one heterogeneous programming language (CUDA/OpenCL/OpenGL).
• Familiar with common deep learning models and algorithms across x86/ARM CPUs, GPUs and embedded chips with platform tuning.
• CV/image processing transplantation experience on diverse hardware; exposure to compilers such as TVM, Glow, LLVM is a plus.
• Strong interest in high-performance and heterogeneous computing, and performance optimization.

Responsibilities

• Collaborate with deep learning engineers to realize software/hardware joint optimization and acceleration for computer vision and speech models on CPU, GPU, DSP, and embedded chips.
• Complete architecture optimization plans for image and ML algorithms on embedded devices and supervise final product-level optimization results.
• Design and implement high-performance algorithms and cross-platform scheduling to enable one-click deployment of models to specific computing platforms.

Skills

C/C++ development

Heterogeneous programming

Tools

CUDA

OpenCL

OpenGL

TVM

Glow

LLVM

Job description

About the job Heterogeneous Computing Expert

1. HeterogeneousComputing Expert

Responsibilities

• Responsible for cooperating with deep learning algorithm engineers to realizethe software and hardware joint optimization and acceleration of computer vision/speechsemantics and other deep learning algorithms on computing platforms such asCPU, GPU, DSP, and specific embedded reasoning chips;
• Complete the architecture optimization plan of the image algorithm and machinelearning algorithm on the embedded device side, and be responsible for thefinal result of the algorithm optimization on the product side;
• Design and implement algorithms with high-performance basic components onend-side-cloud computing platforms, as well as a cross-platform schedulingframework, to achieve one-click deployment of algorithm models to specificcomputing platforms;

Requirements

• Solid computer foundation, familiar with computer architecture and basictechnology of parallel computing;
• More than 5 years of C/C++ development experience, proficient in at least oneheterogeneous programming language, including but not limited to CUDA, OpenCL,OpenGL and other heterogeneous programming;
• Familiar with common deep learning models andalgorithms, with x86CPU, ARM CPU, GPU, DSP and other embedded reasoning chipsand other platform algorithm tuning experience;
• Computer vision and image processing algorithm transplantation experience invarious hardware devices is preferred, and experience in the development ofTVM, glow, LLVM and other compilers is preferred;
• Have a strong interest in high-performance computing, heterogeneous computingand performance optimization.