Principal AI Performance Engineer

Advanced Micro Devices, Inc•San Jose, CA

2d•Hybrid

About The Position

AMD is looking for a performance-obsessed engineer to drive AI inference performance to the absolute limit on AMD GPUs. You will lead a small, highly technical team and work end-to-end across the stack: profiling, diagnosing, and optimizing leading models on customer-relevant serving configurations (e.g. agentic coding, long-context, high-throughput serving). You move from challenge to challenge, tackling the hardest performance problems across our most strategic customer engagements and leaving behind measurable uplifts and reusable methodology. This is not a sustaining role: every engagement is different, every optimization leaves a lasting impact. THE PERSON: You can take any AI workload, understand it top to bottom, and make it faster. You are equally comfortable profiling a distributed serving deployment, diagnosing a kernel-level bottleneck, and presenting optimization results to a customer's VP of Engineering. You understand GPU kernel performance deeply: not just how to use profiling tools, but how to reason about occupancy, cache behavior, memory coalescing, and instruction-level bottlenecks from first principles. You lead through technical depth: you set the standard for your team by doing the hardest work yourself and pulling others up along the way. You are AI-fluent, not just in the models you optimize, but in how you work: you leverage AI agents and tools daily to accelerate your workflows, and you actively define new ways of using them to make yourself and your team more effective. You thrive under pressure, move fast, and measure everything.

Requirements

7+ years of software development experience in GPU computing, AI systems, or high-performance computing
Deep hands-on experience with AI serving frameworks (vLLM, SGLang, TensorRT-LLM, or similar) and their internals
Strong background in end-to-end workload profiling and bottleneck diagnosis: you can trace from user request to GPU kernel and back
Understanding of GPU kernel performance characteristics: occupancy, register and LDS pressure, memory coalescing, cache utilization, wavefront scheduling, and instruction-level bottlenecks
Ability to read and reason about kernel-level profiling data and translate it into concrete optimization actions. You may not write kernels from scratch daily, but you can tell exactly why one is slow and what needs to change
Understanding of model architectures (transformers, MoE, diffusion), inference paradigms (speculative decoding, prefill-decode disaggregation, continuous batching), and how they map to hardware
Experience with custom kernel development or integration (HIP, CUDA, Triton, CK, or similar)
Understanding of multi-GPU and multi-node distributed systems: scale-up and scale-out topologies, RCCL/NCCL, RDMA, and communication-compute overlap
System and rack-level design awareness: understanding performance tradeoffs across the full deployment stack
Strong proficiency in Python and C++
Customer-facing technical leadership experience: ability to engage with customers, present findings, and drive decisions
Fluent in AI-assisted development: daily user of AI agents and tools, with a mindset toward defining new AI-powered workflows
Strong Linux systems knowledge
Excellent written and verbal English communication skills

Responsibilities

Drive performance optimization end-to-end across the stack on leading models and customer-relevant serving configurations, closing competitive gaps through kernel and systems-level optimizations
Profile, diagnose, and resolve the hardest cross-stack performance bottlenecks, from GPU kernels and operator dispatch to framework-level scheduling and multi-node communication
Diagnose kernel-level performance issues using profiling tools: identify occupancy limitations, L2 cache thrashing, register pressure, memory coalescing issues, etc, and translate findings into actionable optimizations
Lead customer-facing technical engagements: present findings, recommend optimizations, and deliver measurable performance uplifts
Integrate and optimize custom kernels (Triton, Gluon, CK, PyDSL, ASM, AITER) within serving frameworks, understanding dispatch paths, shape extraction, and backend selection
Optimize multi-node distributed inference: communication-compute overlap, parallelism strategies, and scale-out performance
Develop and refine shared performance optimization methodology that raises the bar across the broader team
Leverage AI agents to accelerate daily work and define best practices for AI-assisted performance engineering
Upstream optimizations into open-source frameworks such as vLLM, SGLang, and PyTorch