Advanced Engineer - High-Efficiency AI Computing

About the Company

We collaborate with a globally recognized leader in ICT infrastructure and smart devices, delivering comprehensive solutions spanning products and services for carriers, enterprises, governments, and individual users worldwide.

Role Overview

Join as an Advanced Engineer specializing in High-Efficiency AI Computing, where you will innovate at the intersection of algorithms and hardware, focusing on next-generation AI accelerators. Your efforts will significantly advance energy efficiency and compute capability for large-scale AI models by refining microarchitectural features through low-precision and sparse computation techniques.

Primary Responsibilities

• Lead research initiatives on quantization techniques for sophisticated low-precision data formats to extend and enhance our computational ecosystem.
• Design and implement scalable, high-performance kernels such as GEMM and FlashAttention that exploit the hardware strengths of proprietary AI accelerators.
• Identify and resolve systemic performance limitations to propel architectural advancements in forthcoming AI chip designs, optimizing for energy use and throughput.
• Collaborate intimately with integrated circuit design teams to generate detailed technical documents, perform rigorous benchmarking, and contribute high-impact patent ideas.

Required Qualifications

• Ph.D. in Computer Science, Electronic Engineering, Automation, or a closely related technical discipline.
• Comprehensive expertise in computer architecture with emphasis on GPU/NPU microarchitecture implementations.
• Proven domain knowledge in low-precision quantization and sparsity algorithm research.

Preferred Qualifications

• Demonstrated success in optimizing inference for Large Language Models or multi-modal AI models, including hands-on kernel development for high-performance computing.
• Experience in architectural design at the system level for AI inference accelerators aimed at high throughput.
• Strong scholarly presence with publications in major conferences such as ISCA, MICRO, HPCA, ASPLOS, NeurIPS, and CVPR.

Additional Information

Applicants must be prepared to provide personal data and CV for processing according to the applicable privacy policies. Note: only selected candidates will be contacted for further process.