Edge AI Compiler Engineer for Custom Silicon

Amazon; Vancouver; On-site; CAD 115,000 - 192,000; Full time

Job summary

Amazon Development Centre Canada ULC in Vancouver, BC seeks an experienced software engineer to design and implement software for a deep learning accelerator. You will collaborate across hardware, software, and applied science teams to bring models to edge silicon and productize them on consumer devices.

You will also work on compiler infrastructure, model quantization and compression techniques to enable efficient execution on specialized hardware, mentoring juniors and contributing to

Qualifications

• 2+ years design or architecture experience for new/existing systems.
• 3+ years non-internship software development experience.
• 3+ years programming in a modern language (Java/C++/C#) with OO design.
• Experience deploying LLMs in production on AI acceleration hardware.

Responsibilities

• Design and develop software stack for deep learning accelerator.
• Develop compiler passes for graph ingestions, optimizations and partitioning.
• Develop backend code generation across heterogeneous platforms.
• Profile and optimize system performance; build tools as needed.
• Participate in design reviews, API development and documentation.
• Collaborate with hardware, software, applied science and product teams to power more experiences on the DL accelerator.
• Mentor and guide junior engineers.

Skills

2+ years design experience

3+ years software development

OO design in Java/C++/C#

LLMs in production on GPUs/TPU

Education

Bachelor's degree in computer science or equivalent

Tools

C/C++ embedded

Compiler development

Job description

Amazon Development Centre Canada ULC in Vancouver, BC seeks an experienced software engineer to design and implement software for a deep learning accelerator. You will collaborate across hardware, software, and applied science teams to bring models to edge silicon and productize them on consumer devices.

You will also work on compiler infrastructure, model quantization and compression techniques to enable efficient execution on specialized hardware, mentoring juniors and contributing to