Performance Architect

Oho Group; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

OHO Group in San Francisco seeks a Principal Performance Modeling Engineer to turn complex architecture questions into clear, defensible design decisions.

You’ll build analytical and simulation-backed models of GPUs and accelerators, drive silicon and system choices, and predict throughput and latency across AI workloads.

The role requires strong Python and deep knowledge of GPU memory hierarchies, with a track record of models that influenced real engineering decisions.

Qualifications

• Substantial experience in performance modeling, computer architecture or systems performance.
• Ownership of models that influenced real engineering decisions.
• Strong Python skills and understanding of GPU memory hierarchies and distributed AI workloads.

Responsibilities

• Build and own analytical and simulation-backed models of GPU and accelerator systems.
• Drive silicon and system architecture decisions through your work.
• Predict throughput, latency, utilisation and efficiency across AI workloads.
• Validate predictions against measured behaviour and explain results for architecture decisions.

Skills

Performance modeling

Python programming

GPU architecture knowledge

Job description

Help decide how the next generation of AI compute should be built.

An engineering team working on advanced AI computing is looking for a Principal Performance Modeling Engineer to turn complex architecture questions into clear, defensible design decisions.

You’ll investigate where performance is won or lost across accelerators, memory systems and distributed workloads, with ownership of both the models and the engineering behind them.

What you’ll do:

• Build and own analytical and simulation-backed models of GPU and accelerator systems.
• Drive silicon and system architecture decisions through your work.
• Predict throughput, latency, utilisation and efficiency across AI workloads.
• Evaluate compute, memory and interconnect trade-offs, including how workloads scale across devices.
• Validate predictions against measured behaviour and explain what the results mean for architecture decisions.

What you’ll bring:

• Substantial experience in performance modeling, computer architecture or systems performance.
• Personal ownership of models that influenced real engineering decisions.
• Strong Python skills and a sound understanding of GPU architecture, memory hierarchies and distributed AI workloads.
• The judgment to challenge assumptions, investigate unexpected results and defend your conclusions.