Unconventional AI
System Modeling (Performance Models)
Mountain View, CA I US Remote +2 · Remote · Full-time · Posted Apr 1, 2026
- Location
- Mountain View, CA I US Remote
Mountain View
Remote - Education
- MS/PhD in a quantitative field, or BS with substantial, clear evidence of equivalent research/engineering depth.
About the job
Develop simulation frameworks and physics-based performance models to evaluate and compare AI accelerator architectures and support cross-layer system design and hardware verification.
What you'll do
- Build composable, high-fidelity power, performance, and area estimation tools for AI acceleration architectures.
- Create comparative analyses of candidate architectures and existing implementations.
- Work with other teams to define modeling and simulation needs for system design and hardware verification.
What they ask for
- Experience with power profiling, modeling, and simulation for AI workloads.
- Deep understanding of spatial architectures, data orchestration, and dataflow strategies and their tradeoffs.
- Strong experience implementing and debugging ML models in PyTorch or similar, including profiling and optimizing large-scale ML systems.
- Familiarity with hardware accelerator design tools and existing systolic-array accelerator architectures.
Nice to have
- Strong Python engineering skills; experience with PyTorch internals, compilation flows, CUDA, Triton, or other GPU programming approaches.
- Familiarity with analog dynamic systems, efficiency techniques, or building simulation and modeling frameworks.
Tools
TimLoop, Accelergy, NeuroSim, CIMLoop, CACTI, PyTorch, CUDA, Triton
AI focus
Develop and simulate physics-based, energy-efficient computing systems and AI accelerators for machine learning workloads.
Model-written summary · the original posting is authoritative