GPU Performance Architect

Advanced Micro Devices, Inc•Folsom, CA

1d•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: RTG (Radeon Technologies Group) Architecture team is passionate about developing next-generation GPU solutions. As GPU Performance Architect, you will collaborate with a strong architecture, software, and design teams on developing next generation products for data centers and super-computers. You will engage in performance modeling, analysis of ML/HPC workloads, architecture exploration and kernel prototyping. Through your prototyping, experiments and analysis, you will provide valuable insight into new and emerging hardware and software technologies and MI line of products.

Requirements

Excellent analytical and problem-solving skills, along with attention to detail.
Knowledge in GPU architectures, basic knowledge of CPU architecture
Experience in developing high level and cycle accurate performance simulators
Experience in machine learning (ML) networks and frameworks including TensorFlow and PyTorch
Experience in building libraries for key ML operators (e.g., GEMMs, Attention, Collectives).
Understanding of Compute API's such as CUDA, and OpenCL
Background in Network-on-Chip (NoC) design and interconnect systems
Strong programming foundation in C, C++, and scripting languages (Python, etc.)
Bachelors or Masters degree in computer engineering/Electrical Engineering

Nice To Haves

Experience in the compiler stack (graph compilers, MLIR, LLVM, .. etc) is a plus
Experience in hardware modeling and design using RTL or System C is a plus

Responsibilities

Develop high-level and detailed performance models for GPU subsystems, SoCs, and multi-GPU systems.
Build analytical, simulation-based, and workload-driven models to predict application performance.
Generate performance projections for future architectures and product configurations.
Establish performance, latency, throughput, and efficiency KPIs early in product definition.
Prototype ideas across the stack from software, and firmware to architectural models to evaluate and validate ideas.
Collect/summarize data and simulation results for consumption by architects and design teams
Work with architects to propose innovative solutions that can be implemented in next generation HW and SW.
Write and optimize GPU kernels at different levels of abstractions (assembly, Hip, Triton, .. etc).
Write microbenchmarks that can be used for competitive analysis and performance verification.