Fellow AI Compiler Engineer

AMD AMD · Semiconductors · San Jose, CA · Engineering

AMD is seeking an engineering leader for their AI team to enhance compiler infrastructure for data center GPU platforms, focusing on AI applications. The role involves setting technical direction, optimizing performance, mentoring staff, and contributing to the open-source LLVM community. Responsibilities include designing and implementing LLVM backend passes and new code generation infrastructure, partnering with hardware and software teams, and ensuring world-class performance, correctness, and stability of the compiler toolchain.

What you'd actually do

  1. Part of AMD’s AI team focused on improving the compiler infrastructure of current products and developing new groundbreaking AMD technologies in the broader compiler space
  2. Design, implement, and tune sophisticated LLVM backend passes, focusing on low-level optimizations like instruction selection, register allocation, instruction scheduling, software pipelining, and memory access transformations.
  3. Drive the design and implementation of new code generation infrastructure, potentially including custom LLVM components for new or evolving hardware.
  4. Partner closely with Hardware Design, Architecture, Software Runtime, and Kernel/Driver teams on hardware/software co-design efforts to maximize the performance of the overall stack.
  5. Represent the company in the open-source LLVM community, contributing patches, participating in design discussions, and driving key feature integration upstream where applicable.

Skills

Required

  • code generation
  • optimization
  • LLVM backend
  • computer architecture
  • software engineering practices
  • compiler development
  • LLVM IR
  • Machine IR (MIR)
  • SelectionDAG
  • GlobalISel
  • Register Allocation
  • TableGen
  • Instruction Scheduling

Nice to have

  • Ph.D. in Computer Science, Computer Engineering, or a related field
  • Active contribution history to the open-source LLVM project

What the JD emphasized

  • unlock the performance of our data center GPU platforms for AI applications
  • delivering industry-leading compiler optimizations
  • compiler infrastructure that others can build on top
  • setting the technical direction
  • driving the highest-impact optimizations
  • world-class performance, correctness, and stability
  • critical role that directly impacts the performance ceiling of our entire platform
  • exceptional ability to translate complex architectural features into world-class backend implementations
  • maximize the performance of the overall stack

Other signals

  • AI applications
  • compiler infrastructure
  • performance ceiling
  • code generation
  • LLVM backend