Principal Software Development Eng. - AI Performance
About The Position
AMD is looking for a Principal Engineer to serve as a hands-on technical team lead driving the performance and scalability of frontier AI workloads on AMD GPUs, including large language models, mixture-of-experts architectures, and diffusion models. You will lead a team of engineers, define the long-term technical vision, make critical architecture decisions, and tackle the hardest performance challenges across the stack from GPU kernels and to serving frameworks and distributed systems. THE PERSON: The ideal candidate is a deep technical expert with a track record of solving industry-hard problems at the intersection of GPU architecture, AI systems, and high-performance software. You understand the full stack from hardware micro-architecture to model architecture, inference paradigms, and system-level design. You lead through technical depth, influence, and by example, staying hands-on while setting direction for your team. If you want to shape how the world runs AI on AMD hardware, this role is for you.
Requirements
- 10+ years of software development experience in GPU computing, HPC, or AI systems
- Deep understanding of GPU micro-architecture, memory hierarchy, instruction scheduling, and performance tradeoffs
- Deep understanding of end-to-end AI systems: model architectures, inference paradigms, and system/rack-level design
- Understanding of multi-GPU communication: scale-up (NVLink, xGMI, Infinity Fabric) and scale-out (RDMA, RCCL/NCCL) topologies and performance characteristics
- Experience designing and optimizing across the full stack: from low-level GPU kernels to frameworks and distributed serving systems
- Strong background in performance engineering, including profiling, roofline analysis, and bottleneck diagnosis at scale
- Experience with one or more of: HIP, CUDA, OpenCL, Triton/Gluon, CUTLASS, CK
- Strong proficiency in C++ (C++17 or later) and Python
- Experience leading small technical teams while remaining a hands-on contributor
- Track record of influencing technical direction across teams and organizations
- Strong Linux systems knowledge
- Excellent written and verbal English communication skills
Nice To Haves
- Experience with GPU compiler toolchains (e.g., LLVM) and intermediate representations (e.g., MLIR, LLVM IR, Triton IR) is a plus
- Hands-on experience contributing to or architecting major open-source AI frameworks (e.g., vLLM, SGLang, xDiT, Megatron LM, PyTorch)
- Published research or significant open-source contributions in GPU computing, HPC, or AI systems is a plus
Responsibilities
- Lead a small team of engineers: set technical direction, prioritize work, and ensure delivery while remaining deeply hands-on
- Define and drive the long-term technical strategy for AI workload performance on AMD GPUs
- Own the most complex cross-stack performance challenges, from kernel optimization to framework-level architecture decisions
- Lead the design and implementation of novel GPU kernels, compiler optimizations, and framework features
- Establish performance methodology and roofline analysis practices that set the standard for the team
- Influence upstream roadmaps in major open-source AI frameworks (e.g., vLLM, SGLang, PyTorch)
- Drive architecture decisions for emerging inference paradigms (e.g., prefill-decode disaggregation, speculative decoding, distributed serving)
- Identify and close fundamental performance gaps between AMD and competitor platforms
- Serve as a technical authority across the organization, advising leadership on technical direction and feasibility
- Mentor engineers and raise the technical bar across the broader engineering organization
- Represent AMD externally through publications, conference talks, and open-source contributions
Benefits
- AMD benefits at a glance.
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Job Type
Full-time
Career Level
Principal
Education Level
Ph.D. or professional degree
