PyTorch Integration Engineer for Accelerated ML

Graphcore; West of England; Hybrid; GBP 85,000 - 140,000; Full time

Benefits offered by this job

Flexible working

Generous leave

Retirement planning support

Phantom equity

On-site barista and free food

Income protection and life assurance

Private medical insurance and dental

Cycle to work

Give as you earn

Job summary

Graphcore is seeking a Software Engineer to help make Graphcore hardware feel native inside ML frameworks. You will design, implement, and optimize features that connect Graphcore accelerators with PyTorch, Triton and related technologies.

Join a SCRUM team in Frameworks, contribute to upstream PyTorch, and support a hybrid Python/C++ extension. You will solve real engineering problems, ship tooling, and improve performance and developer experience for ML researchers.

Qualifications

• Strong core software engineering skills and a collaborative mindset.
• Experience developing Python libraries.
• Experience building performant C++ applications in a commercial setting.
• Understanding of linear algebra, floating-point arithmetic or numerical analysis.
• Confident using Git and code reviews in a commercial team.
• Willingness to explore large, unfamiliar open-source codebases independently and with others.

Responsibilities

• Design, implement, optimise and support PyTorch-related features across PyTorch, Triton and related technologies.
• Contribute to upstream PyTorch and support a hybrid Python and C++ extension.
• Help integrate compiler backend capabilities with torch.compile and improve performance and developer experience.

Skills

Strong core software engineering

Experience developing Python libraries

Experience building performant C++

Understanding of linear algebra / FP /

Git proficiency

Open-source experience

Job description

Graphcore is seeking a Software Engineer to help make Graphcore hardware feel native inside ML frameworks. You will design, implement, and optimize features that connect Graphcore accelerators with PyTorch, Triton and related technologies.

Join a SCRUM team in Frameworks, contribute to upstream PyTorch, and support a hybrid Python/C++ extension. You will solve real engineering problems, ship tooling, and improve performance and developer experience for ML researchers.