AI Performance Library Architect

About the role

We are seeking a software development engineer to join the oneDNN project (https://github.com/uxlfoundation/oneDNN), a complex cross-platform open-source software project focusing on neural network performance. oneDNN is a critical and highly visible component of Intel’s AI strategy, powering key AI applications including OpenVINO, TensorFlow, PyTorch, and ONNX Runtime.

In this role, you will design, develop, and maintain new functionality in oneDNN to enable performance-critical portions of AI workloads. You will support software developers optimizing AI frameworks and workloads for Intel CPUs and GPUs, as well as a cross-platform ecosystem of AI software developers contributing to oneDNN.

Responsibilities

• Design, develop, and maintain new functionality in oneDNN.
• Enable performance-critical portions of AI workloads.
• Support software developers optimizing AI frameworks and workloads for Intel CPUs and GPUs.
• Engage with the cross-platform ecosystem of AI software developers contributing to oneDNN.

Requirements

Minimum qualifications:

• Master’s degree or Ph.D. in Mathematics, Physics, Computer Science, or a relevant STEM field.
• 5+ years of experience in the following areas:
  • C and C++
  • Maintaining or contributing to open-source software projects
  • Software libraries design and architecture
  • Implementation of linear algebra algorithms (functions from BLAS, LAPACK, or PyTorch)
  • Performance engineering and software performance optimizations
  • Floating point arithmetic and numerical stability
  • Software development on Linux
  • Low-level performance optimizations using CUDA, x86 assembly or intrinsics, or OpenCL

Preferred Qualifications

• 3+ years of experience in machine learning and deep learning algorithms or high-performance computing (HPC) applications development.
• 3+ years of experience with floating point implementations of transcendental functions (sin, cos, tanh, elu, etc.).
• 1+ years of experience with algorithms for non-IEEE low precision data types (bfloat16, fp8, fp4).
• 1+ years of experience with AI-assisted software development.

Requirements may be obtained through a combination of industry-relevant job experience, internships, and/or schoolwork, classes, or research.

Benefits

We offer a total compensation package that ranks among the best in the industry, including:

• Competitive pay and stock bonuses.
• Comprehensive benefits programs covering health, retirement, and vacation.

Pay

Annual salary range for roles that could be performed in the US: $170,500 – $315,490 USD. Individual pay is determined by job-related skills, experience, relevant education or training, and work location.

Schedule

This role follows a hybrid work model, allowing employees to split their time between working on-site at their assigned Intel site and off-site.

Job details, including work model, location, or time type, are subject to change.