MLIR/LLVM Compiler Engineer for Hexagon DSP

QUALCOMM, Inc.; Austin (TX); Sur place; USD 117 000 - 175 000; Plein temps

Résumé du poste

Qualcomm Technologies, Inc. in Austin, TX is seeking compiler engineers to advance LLVM/MLIR-based components for the Hexagon DSP, enabling high-performance ML, signal processing, and multimedia workloads across Qualcomm devices.

You will collaborate with hardware and software teams to optimize code generation and performance on our silicon. You will work on the full compilation stack, including lowering, analyses, and debugging across MLIR, LLVM IR, and assembly, while scaling ML frameworks for

Qualifications

• Hands-on experience with LLVM/MLIR or GCC in production compilers.
• Strong C/C++ programming in large software projects.
• 5+ years of compiler engineering experience in industry or academia.
• Bachelor’s degree in CS/CE/ECE or related field.

Responsabilités

• Develop LLVM- and MLIR-based compiler components for Qualcomm’s Hexagon DSP.
• Improve ML compiler lowering and optimization pipelines for efficient execution on Qualcomm hardware.
• Use profiling and benchmark data to drive performance and code size improvements.
• Debug correctness and code generation across MLIR, LLVM IR, assembly, and generated code.
• Collaborate with architecture and HW engineers to enable efficient silicon usage.

Connaissances

LLVM MLIR

C/C++

Compiler engineering

Large codebases

ML compiler stacks

Formation

Bachelor's degree in CS/EE

Master's degree

PhD

Outils

Assemblers

Linkers

Debuggers

Profilers

Description du poste

Qualcomm Technologies, Inc. in Austin, TX is seeking compiler engineers to advance LLVM/MLIR-based components for the Hexagon DSP, enabling high-performance ML, signal processing, and multimedia workloads across Qualcomm devices.

You will collaborate with hardware and software teams to optimize code generation and performance on our silicon. You will work on the full compilation stack, including lowering, analyses, and debugging across MLIR, LLVM IR, and assembly, while scaling ML frameworks for