Machine Learning Hardware/Software Codesign Engineer
AkeanaRole Summary
The Machine Learning Hardware/Software Codesign Engineer will be a key player in designing and optimizing machine learning accelerators alongside software infrastructure in a startup environment.
Experience Level
Mid-level to Senior position; a Master’s or PhD or equivalent experience in a relevant discipline (Computer Science, Computer Science & Engineering, or Computer Engineering) is required.
Responsibilities
Key responsibilities include:
- Developing software architecture and APIs for high-performance extensible software.
- Conducting performance, power, and accuracy analyses on various deep learning techniques.
- Optimizing deep learning models, particularly for on-device inference.
- Collaborating with team members on performance analysis of hardware and software stacks.
- Prototyping innovative solutions to improve inference performance.
Requirements
Must-have skills include:
- Strong background in Python/PyTorch with hands-on machine learning experience.
- Proficiency in C/C++ programming and software design.
- Knowledge of computer architecture (CPU and GPU) and performance modeling.
- Experience with ML systems for on-device inference.
- Excellent communication and collaboration skills.
Nice-to-have skills include:
- Experience moving models from PyTorch into MLIR.
- Familiarity with Triton and TensorRT.
Education Requirements
Master’s or PhD or equivalent experience in a relevant discipline (Computer Science, Computer Science & Engineering, or Computer Engineering).
About the Company
Company: Akeana
Headquarters: Santa Clara, CA, USA
Akeana is a start-up specializing in processor design, with a team experienced in developing state-of-the-art products. Their expertise spans multiple architectures, including Risc-V, ARM, and x86, with backgrounds from prominent companies in the semiconductor industry. Funded by top-tier venture capitalists and industry leaders, Akeana focuses on innovative RTL design and micro-architecture development.
