Compute Library Engineer

Full Job Description

Compute Library Engineer in Cambridge, UK

About Neubla

As pioneers of innovative neural computing, Neubla is transforming the future of computing by bringing the enormous potential of neural computing to the highest performance in most data-intensive applications. Neubla is made up of people with expertise and lots of experience in many different fields such as AI, semiconductor design, and compilers. At Neubla, you can work in a rapidly changing environment where your new ideas will become innovative products, services, and customer experiences, working with talented colleagues. Neubla offers a competitive package with benefits, including health insurance, a pension scheme, employee wellness support, and a variety of snacks in the office!

Summary

As a Compute Library (CL) Engineer at Neubla, you will play a crucial role in the dynamic team responsible for crafting and optimizing the Compute Library tailored specifically for our designated hardware platform. Your contributions will be instrumental in advancing the capabilities and performance of our cutting-edge technology.

Job Description

• Develop high-performance kernels for machine learning operators on NPU architectures
• Optimize existing kernels to ensure maximum efficiency and performance
• Collaboratae with the hardware and software teams to integrate kernels into the NPU framework
• Conduct thorough testing and debugging of kernels to ensure reliability and functionality
• Utilize hardware features of GPU and accelerators specialized for AI applications, including convolutions, matrix multiplication units and high-bandwidth interconnections
• Document and maintain clear, concise, and comprehensive code and design documentation

Qualifications

• Masters, PhD, or equivalent experience in Computer Engineering, Electrical Engineering, or related field
• Minimum of 5 years of experience in kernel development for NPUs, GPUs, or similar hardware accelerators from 3 ~ 5 years (ML, linear algebra)
• Proficiency in CUDA, OpenCL, or similar parallel programming languages
• Experience in SIMD/vector processing experience(RISC-V Vector)
• Strong software development skills using standard development tools (e.g., Git, Jira, etc.)
• Basic understanding of machine learning frameworks (TensorFlow, PyTorch, etc.)
• Excellent problem-solving skills with a logical and analytical mindset
• Strong communication and interpersonal skills to collaborate effectively within a team environment
• Ability to work independently, manage multiple tasks, and thrive in a fast-paced, dynamic work environment

Preferred Qualifications

• Experience in optimizing kernels for power efficiency and low-latency applications
• Familiarity with hardware architecture and system-level understanding of NPUs or GPUs
• Knowledge of performance profiling tools and methodologies
• Contributions to projects related to parallel computing or machine learning

The Process

• Application review > 1st Interview > 2nd Interview (technical) > 3rd Interview