AI Performance Modeling Engineer

Job Title

AI Performance Modeling Engineer

Role Summary

Model and project the performance of generative AI and large-model workloads on AMD GPU architectures before silicon is available. Work spans analytical performance bounds, architectural simulation, and cycle‑accurate simulation to influence GPU microarchitecture and software stacks.

Collaborate with architecture, RTL, compiler, runtime, and framework teams to translate workload behavior into concrete hardware recommendations and verify projections on early silicon.

Experience Level

Mid-level (no explicit years-of-experience specified in the posting).

Responsibilities

Primary duties focus on projecting AI workload performance, diagnosing bottlenecks, and driving architecture trade-offs.

• Project AI/LLM performance using analytical models, architectural simulation, and cycle-accurate simulators.
• Select appropriate model fidelity (rooflines/analytical for trade-offs; detailed simulation for precision).
• Identify and quantify performance limits across compute, cache hierarchies, memory bandwidth, and SoC interconnect.
• Characterize workloads via tracing and profiling; extract footprints and map them to microarchitectural constraints.
• Run architecture experiments and sensitivity analyses to evaluate GPU configurations and scaling behavior against real AI workloads.
• Extend and maintain simulation infrastructure for compute engines, caches, memory subsystems, and interconnects.
• Validate models on early silicon, correlate telemetry to pre-silicon projections, and root-cause mismatches using low-level counters and traces.
• Partner with compiler/runtime/framework teams to convert model insights into optimizations and architectural requirements.

Requirements

Concise list of must-have technical skills and one item of preferred experience.

• Must-have: Deep, hands-on understanding of LLM/Transformer performance, including scaling techniques and tensor/pipeline/data parallelism.
• Must-have: Practical experience building and using analytical performance models and execution-driven / architectural / cycle‑accurate simulators; able to choose appropriate fidelity.
• Must-have: Strong GPU architecture knowledge: execution pipelines, SIMD/SIMT, cache hierarchies, memory technologies, and high-bandwidth interconnects.
• Must-have: Production-grade programming skills in modern C++ and Python.
• Must-have: Strong structured problem-solving and ability to turn simulation data into actionable recommendations.
• Nice-to-have: Experience with NPU / AI-accelerator architecture or modeling on dedicated inference/training accelerators.

Education Requirements

Not specified.

About the Company

Company: Advanced Micro Devices

Headquarters: Sunnyvale, California, USA

Advanced Micro Devices, or AMD, is a global semiconductor company that designs and manufactures microprocessors, graphics processors, and related technologies for a variety of computing devices. Known for pushing the boundaries of innovation, AMD's mission is to deliver high-performance computing solutions for AI, data centers, gaming, and embedded applications. They foster a collaborative, inclusive culture focused on creativity and problem-solving, aiming to drive progress and excellence in technology.