AI Performance Modeling Engineer
Core
Model AI workload performance on GPU architecture to predict speed before silicon exists and shape hardware design.
Role type
Senior IC AI performance modeling engineer
Builds
GPU microarchitecture for AI workloads (Transformers, LLMs, vision models)
Domain
Semiconductor / GPU Architecture / AI Systems
Deliverable
production ML models
Required skills
Analytical performance modeling, Cycle-accurate simulation, GPU architecture (SIMD/SIMT, cache hierarchy, memory subsystems), Modern C++, Python, LLM/Transformer scaling strategies, Roofline analysis, Arithmetic intensity, Tensor/pipeline/data parallelism, Kernel execution graphs
Preferred skills
NPU / AI-accelerator architecture modeling
Technologies
C++, Python, Cycle-accurate simulators, Performance counters, Linux tracing
Responsibilities
Build analytical and cycle-accurate models to project AI workload performance; Characterize AI workloads to identify microarchitectural bottlenecks; Extend and maintain GPU subsystem simulators; Correlate pre-silicon models with post-silicon telemetry; Partner with compiler and runtime teams to translate insights into architectural constraints.
