Deep Learning Compiler Engineer (MLIR/LLVM)

Job Description:

Join our compiler team and contribute to the development of an MLIR-based compiler that drives performance improvements on Intel deep learning accelerators.

This compiler delivers significant performance gains across Intel products and directly impacts cutting-edge deep learning workloads.

In this role, you will:

• Design and implement new optimizations within the MLIR and LLVM frameworks to enhance model-level performance for deep learning applications.
• Collaborate with architecture and performance teams to identify and address bottlenecks in the compiler pipeline.
• Engage with internal customers and developers to understand requirements and support model-level performance tuning.
• Explore and prototype novel compilation techniques to improve hardware utilization and efficiency.

This position is expected to change location in the near future to the Petah Tikva campus.

Qualifications:

Required:

• 5+ years of experience in C++ and familiarity with modern software development practices.
• Background in high-performance computing or parallel programming.
• Strong analytical and problem-solving skills.
• Experience with development on Linux.

Advantage:

• Strong experience with compiler design and middle-end optimizations, preferably using MLIR and LLVM.
• Experience with code generation, performance analysis, and tuning for hardware accelerators.
• Knowledge of dataflow graphs, IR transformation, and optimization techniques.