Kernel Engineer

Job Title

Kernel Engineer

Role Summary

Develop high-performance software at the intersection of hardware and software to implement, optimize, and scale deep learning and HPC kernels for a custom massively-parallel processor.

Work on a team responsible for designing, performance-tuning, validating, and maintaining a library of parallel and distributed algorithms that maximize compute utilization on Cerebras hardware.

Experience Level

Mid-level. The posting does not specify exact years of experience.

Responsibilities

Design, implement, optimize, and validate ML and linear-algebra kernel routines to run efficiently on Cerebras hardware.

• Create design specifications and map algorithms to the Cerebras WSE system using parallel programming techniques.
• Implement and debug low-level kernel routines in assembly and a C-like domain-specific language (CSL).
• Use mathematical models and performance analysis to guide design and optimization decisions.
• Develop and integrate unit and system tests to verify functionality and performance of kernel libraries.
• Collaborate with chip and system architects to optimize instruction sets, microarchitecture, and I/O.
• Research emerging ML application trends and evolve kernel library architecture to meet computational needs.

Requirements

Must-have technical skills and abilities.

• Comfort with learning and working close to hardware architecture details.
• Proficiency in C++ and Python.
• Experience or strong knowledge of library and API development best practices.
• Strong debugging skills for complex software stacks.
• Ability to perform performance analysis and use mathematical models to assess software behavior.
• Experience writing tests and integrating verification methodologies for performance- and correctness-critical code.
• Effective collaboration skills with architects and cross-functional engineering teams.

Education Requirements

Bachelor's, Master's, PhD or foreign equivalents in Computer Science, Computer Engineering, Mathematics, or related fields.

About the Company

Company: Cerebras

Headquarters: Sunnyvale, CA, USA

Developer of wafer-scale AI accelerators, Cerebras designs the Wafer Scale Engine (WSE)—one of the world’s largest AI chips—to deliver high-speed training and inference solutions for model labs, enterprises, and AI-native startups.