Senior Software Engineer – AI Compute Libraries & Performance

Senior Software Engineer – AI Compute Libraries & Performance Overview

As a Senior Software Engineer specializing in AI Compute Libraries and Performance, you will be tasked with developing high-performance compute kernels that maximize the capabilities of next-generation AI hardware. Your role will involve close collaboration with hardware, where every design decision is crucial. You will be responsible for creating kernels for various linear algebra and tensor operations, including GEMM, convolutions, reductions, and more. Your focus will be on enhancing performance, ensuring correctness, and maintaining numerical reliability across essential AI workloads.

You will join the ML Kernels and Runtime team, which is dedicated to high-performance compute libraries and operates closely with both hardware and product development. The team values clear ownership, practical decision-making, and rapid technical feedback. As part of this team, you will be expected to contribute to the product’s profiling, testing, debugging, and overall improvement while being accountable for tangible outcomes. Mentoring and sharing knowledge with colleagues is also a key aspect of this role.

Responsibilities and Duties

• Design and implement kernels for linear algebra and tensor operations in C++.
• Own performance and correctness through microbenchmarks, regression tests, and numerical validation.
• Profile and optimize performance for next-generation AI hardware.
• Debug issues and enhance product quality and functionality.
• Engage in Agile practices within the team.
• Mentor team members and share your expertise.

Candidate Profile

Essential Qualifications

• Excellent programming and scripting skills in C++ and Python.
• Understanding of processor architectures and experience profiling on Linux.
• Strong written and verbal communication skills, with a collaborative work ethic.
• Commitment to producing quality work and being a team player.

Desirable Qualifications

• Strong grasp of algorithmic performance, including vectorization and memory hierarchy.
• Experience with at least one BLAS/DNN stack and the ability to read and extend kernels.
• Comfort with CPU micro-optimizations and understanding of numerical stability across various floating-point formats.
• Experience integrating native code into PyTorch or similar frameworks.
• Knowledge of ABI/API stability and Linux system packaging.

Graphcore is committed to fostering an inclusive environment where diverse backgrounds and experiences are valued. We encourage applicants from all walks of life and are dedicated to ensuring that everyone can perform at their best. If you require any adjustments during the interview process, please let us know, and we will be happy to accommodate you.

If you are eager to contribute to the development of compute libraries for next-generation AI hardware, we invite you to apply and be part of our innovative team.

Degree Requirement: Not Specified

Visa Sponsorship May be