Performance Modeling Lead

Oho Group; San Francisco (CA); On-site; USD 180,000 - 320,000; Full time

Job summary

Oho Group is seeking a Principal Performance Modeling Architect to lead the modelling platform that directly shapes silicon design, cluster architecture and commercial decisions. You’ll own first-principles models of performance, energy, and cost across accelerators and clusters.

You will extend modelling to LLM inference, training and multimodal workloads, analyzing compute, memory and interconnect trade-offs and validating predictions against real frameworks.

Qualifications

• Strong experience in performance modelling, computer architecture or systems performance, with models used to influence decisions.
• Strong Python, analytical/roofline modelling and an understanding of GPU hardware and distributed AI workloads are essential.
• Understanding of both inference and training workloads.

Responsibilities

• Build first-principles models of performance, energy and cost across accelerators and clusters.
• Extend modelling across LLM inference, training and multimodal workloads.
• Analyse compute, memory and interconnect trade-offs, validating predictions against real frameworks.
• Own a production Python platform and develop AI-assisted analysis capabilities.

Skills

Performance modelling

Computer architecture

GPU/AI workloads

Analytical modelling

Tools

Python (data analysis)

Job description

Principal Performance Modeling Architect

Before committing to new silicon, how do you prove which architecture will deliver the best performance?

I’m working with a well-funded GPU and AI silicon startup hiring an architect to answer exactly that. You’ll own the modelling platform that directly shapes silicon design, cluster architecture and commercial decisions.

The work

• Build first-principles models of performance, energy and cost across accelerators and clusters.
• Extend modelling across LLM inference, training and multimodal workloads.
• Analyse compute, memory and interconnect trade-offs, validating predictions against real frameworks.
• Own a production Python platform and develop AI-assisted analysis capabilities.

Your background

• Strong experience in performance modelling, computer architecture or systems performance, with models you’ve built and used to influence real decisions.
• Strong Python, analytical/roofline modelling and an understanding of GPU hardware and distributed AI workloads are essential.
• Understanding of both inference and trainingworkloads.

A hands‑on architecture role with direct influence over what gets built, backed by an experienced semiconductor team and meaningful equity participation.

Keywords

vLLM / SGLang / HPC / TensorRT / FSDP / Accelerator / GPU / Inference / Modeling / Models / XPU / HBM / MCP