Performance Analysis Lead for ML Accelerator Hardware

The Consensus; San Jose (CA); On-site; USD 180,000 - 260,000; Full time

Benefits offered by this job

Medical coverage

Housing subsidy

Relocation support

Wellness benefits

Lunch provided

Unlimited compute budget

Job summary

Etched, building hardware for frontier intelligence, is seeking a performance analysis engineer to design a tooling suite for its ML accelerator. You will enable ML engineers and customers to understand workload behavior, identify bottlenecks, and unlock full hardware potential across hosts, PCIe, and accelerators.

Collaborate with architects, firmware, drivers, compilers, and ML teams to define requirements and deliver intuitive analyses and visualizations that communicate performance

Qualifications

• Strong proficiency in C++ or Rust.
• Python proficiency is a plus.
• Deep understanding of computer architecture and PCIe interconnects.
• Experience with low-level performance analysis on complex hardware.
• Familiar with performance analysis tools such as Nsight, VTune, and perf.

Responsibilities

• Lead the design and architecture of a performance analysis suite.
• Develop methods to capture performance data from custom hardware.
• Implement tracing for host-side API calls and system events.
• Correlate performance events across host, drivers, PCIe, accelerators, and hosts.
• Build analysis modules to interpret trace and counter data.
• Develop visualizations like timelines and graphs.
• Collaborate with hardware, firmware, driver, compiler, and ML teams.

Skills

C++/Rust

Python

Computer architecture

Performance analysis

Hardware counters

Tools

NVIDIA Nsight

Intel VTune

AMD uProf

perf

Tracy

ETW

Job description

Etched, building hardware for frontier intelligence, is seeking a performance analysis engineer to design a tooling suite for its ML accelerator. You will enable ML engineers and customers to understand workload behavior, identify bottlenecks, and unlock full hardware potential across hosts, PCIe, and accelerators.

Collaborate with architects, firmware, drivers, compilers, and ML teams to define requirements and deliver intuitive analyses and visualizations that communicate performance