AI Inference Systems Performance Engineer

Tensordyne; Sunnyvale (CA); On-site; USD 140,000 - 210,000; Full time

Benefits offered by this job

Comprehensive benefits

Flexible spending options

Recognition programs

Job summary

Tensordyne in Sunnyvale, CA, seeks a Systems Performance Modeling Engineer to build models and tools predicting AI inference workloads on our systems from single accelerators to rack-scale deployments. This hands-on role involves writing simulator code, running experiments, and analyzing discrepancies between predictions and measurements.

You'll collaborate across architecture, silicon, hardware, and software teams to extend models of silicon, interconnects, and fabric, capture workload

Qualifications

• Hands-on experience building performance models, simulators, or analytical tools for ML workloads.
• Strong programming skills in C++ and Python with clean, testable code.
• Experience comparing model predictions against real measurements and debugging divergences.
• Understanding of distributed ML execution and parallelism strategies.

Responsibilities

• Build and extend simulation-based performance models for multimodal AI inference at rack, pod, and cluster scale.
• Model batching, KV-cache management and disaggregation effects on latency and throughput.
• Create trace-capture and replay tooling for hardware configurations.
• Run calibration experiments on Tensordyne hardware and compare to model predictions.
• Provide design-space analyses and performance projections for decisions.

Skills

C++

Python

Distributed systems

Performance modeling

Debugging

Communication

Education

MS or higher in CS/CE/EE or related field

Tools

NCCL

Python testing

Simulation tools

Profiling ML workloads

Job description

Tensordyne in Sunnyvale, CA, seeks a Systems Performance Modeling Engineer to build models and tools predicting AI inference workloads on our systems from single accelerators to rack-scale deployments. This hands-on role involves writing simulator code, running experiments, and analyzing discrepancies between predictions and measurements.

You'll collaborate across architecture, silicon, hardware, and software teams to extend models of silicon, interconnects, and fabric, capture workload