SENIOR SILICON DESIGN ENGINEER: Hardware-Software Co-Design, GPU Compute and AI Compilers & Runtimes

Advanced Micro Devices, Inc•Bellevue, WA

3d•Hybrid

About The Position

At AMD, our mission is to build great products that accelerate next-generation computing experiences—from AI and data centers, to PCs, gaming and embedded systems. Grounded in a culture of innovation and collaboration, we believe real progress comes from bold ideas, human ingenuity and a shared passion to create something extraordinary. When you join AMD, you’ll discover the real differentiator is our culture. We push the limits of innovation to solve the world’s most important challenges—striving for execution excellence, while being direct, humble, collaborative, and inclusive of diverse perspectives. Join us as we shape the future of AI and beyond. Together, we advance your career. THE ROLE: The AMD AI Group is looking for a Senior Silicon Design Engineer to co-design the hardware-software interface for AMD Instinct™ GPUs. You will work with hardware architects to model and evaluate how GPU microarchitecture, AI frameworks, and the ROCm™ software stack interact, using that analysis to drive improvements in both performance and energy efficiency across the AI compute stack. Your day-to-day: build performance and power models that connect architectural decisions to real workload behavior. Evaluate proposed hardware features and software API designs against AI training and inference targets. Write and optimize GPU kernels and compiler passes that exploit microarchitectural characteristics. Work across the stack: the Linux kernel DRM/GEM layer, the ROCm™ runtime, the IREE compiler, and AI frameworks and DSLs like PyTorch and Triton.

Requirements

5+ years of industry experience working at the boundary of GPU compute hardware and AI systems software.
Strong computer architecture foundation: memory consistency and coherence models, GPU microarchitecture, pipelining, and the ability to reason precisely about what hardware is doing on a given workload.
Demonstrated performance modeling skills and the ability to build analytical or simulation-based models that predict real hardware behavior and use them to make engineering decisions.
Proficiency in C and C++ in system-level contexts, and Python for tooling, analysis, and framework-level work.
Experience with GPU programming via CUDA and/or HIP, including writing and optimizing compute kernels beyond library calls.
Familiarity with Linux kernel internals relevant to GPU compute: DRM/GEM subsystem, kernel-mode driver interfaces, memory management, and userspace-kernel interaction patterns.
Experience with compute runtimes and the full dispatch path from framework-level operations to GPU execution.
Experience with AI framework internals including PyTorch dispatcher, Triton compiler, or equivalent depth in how frameworks lower operations to hardware.

Nice To Haves

Direct experience with AMD Instinct GPU architectures (MI300X, MI350X, MI355X) and the ROCm™ stack.
Working knowledge of compiler infrastructure such as LLVM/MLIR-based flows, backend code generation, or related optimization passes. Experience with IREE is a strong plus.
Familiarity with modern AI model architecture: quantization techniques, KV cache, attention mechanisms and their variants, mixture-of-experts routing, or emerging state space models and an understanding of what these architectures demand from hardware.
Experience with performance analysis tools for GPUs like profilers, trace viewers, hardware counters and building custom tooling on top of them.
Publications or patents in HPC, ML systems, or GPU kernel optimization.

Responsibilities

Leverage detailed performance models that bridge GPU microarchitecture and software execution, including instruction execution, memory hierarchy behavior, networking, and compute-memory overlap. You will use those models to drive both software optimization and hardware design feedback.
Analyze and characterize AI workloads (training and inference) against AMD Instinct™ microarchitecture to identify performance bottlenecks, inform silicon roadmap decisions, and validate that architectural features deliver expected real-world speedups.
Collaborate directly with AMD GPU hardware and firmware teams to evaluate proposed hardware features, providing software-grounded analysis of how changes to the memory subsystem, cache coherence protocols, or execution units affect workloads that matter.
Own the feedback loop between workload characterization and hardware design: build the methodology, tooling, and benchmarks that make this loop fast and evidence-driven across Instinct generations.
Contribute to the IREE compiler and ROCm™ stack, with focus on backend optimizations, runtime API extensions, and firmware command packet design that target AMD GPU microarchitecture.
Design and optimize compute runtime paths including AI framework dispatch, data movement, and pipelining, minimizing overhead and maximizing hardware utilization for latency-sensitive and throughput-critical workloads.
Work at the Linux kernel level on GPU compute infrastructure to improve asynchronous execution and reduce latency.
Contribute to the open ROCm™ ecosystem, across the stack, in ways that raises the bar for the entire AMD AI developer experience.
Develop reproducible benchmarks and performance regression infrastructure that keeps the full stack honest across compiler, runtime, and driver changes.
Represent the software perspective in silicon architecture reviews and represent hardware realities to the software and framework teams. Be the person who holds both models in their head at once.
Collaborate with AMD architecture teams to provide software feedback on next-generation Instinct™ GPU designs for both training and inference workloads.