NVIDIA
Senior Software Engineer, Agent Performance Modeling
$152K–$288K
US, CA, Santa Clara · Senior · Full-time · Posted Oct 9, 2026
- Experience
- 5+ years of experience.
- Education
- MS in Computer Science or a related field.
About the job
Builds performance models and agent-driven simulation workflows to evaluate LLM architectures on GPUs, LPUs, and heterogeneous systems, guiding hardware and software decisions.
What you'll do
- Build and validate performance models for LLM architectures across GPU, LPU, and heterogeneous systems.
- Develop agents to automate simulation workflows, analyze bottlenecks, and accelerate design space exploration.
- Collaborate with hardware architecture, inference, runtime, and compiler teams to improve modeling accuracy and guide optimization.
What they ask for
- Familiarity with modern LLM architectures and their inference characteristics.
- Experience with simulation, profiling, or performance modeling, including familiarity with roofline analysis for LLM inference.
- Strong analytical skills and a rigorous approach to experimentation and validation.
- Strong software engineering skills and experience building reliable engineering tools using coding agents like Claude Code, Codex, etc.
Nice to have
- Experience building agents that automate complex engineering workflows.
- Knowledge of accelerator architecture and compute, memory, or communication bottlenecks; open-source contributions in agentic systems, simulation, or inference optimization.
Pay
152,000 USD - 241,500 USD (base salary, Level 3); 184,000 USD - 287,500 USD (base salary, Level 4)
Tools
LLM, GPU, LPU, Claude Code, Codex
AI focus
Develop agents that automate simulation and engineering workflows, analyze performance bottlenecks, and support design space exploration for LLM systems.
Model-written summary · the original posting is authoritative