Software Engineer – Performance Profiling

Job Title

Software Engineer – Performance Profiling

Role Summary

Lead design and implementation of a performance analysis and profiling toolchain for a custom ML accelerator. You will build data collection, tracing, correlation, analysis, and visualization components that help engineers and customers diagnose performance bottlenecks across host and device.

Work closely with hardware architects, firmware, driver, compiler, and ML application teams to define requirements and enable optimizations. This is an onsite role in San Jose, CA.

Experience Level

Mid-level engineer. The posting does not state specific years of experience but expects proven experience in low-level performance analysis and systems programming.

Responsibilities

Deliver end-to-end performance tooling and analyses for a PCIe-attached ML accelerator.

• Design architecture for a performance analysis suite: collection, pipelines, analysis engines, and user interfaces (CLI/GUI).
• Implement low-overhead data collection from hardware (performance counters, execution unit status, memory access) via drivers or services.
• Instrument and trace host-side APIs, runtime libraries, driver interactions, and system events (CPU, PCIe, memory, network).
• Correlate events across CPU, driver, PCIe, multiple devices, and hosts with precise time synchronization.
• Build analysis modules to identify compute, memory, PCIe, and other hardware bottlenecks.
• Create visualizations (timelines, dependency graphs, utilization charts, summaries) to communicate findings.
• Collaborate with cross-functional teams to validate tooling and guide performance optimization.

Requirements

Must-have qualifications and core skills required to perform the role.

• Strong proficiency in C++ or Rust for systems-level tooling.
• Experience with low-level performance analysis and profiling on complex hardware (GPUs, CPUs, FPGAs, or custom accelerators).
• Deep understanding of computer architecture, memory hierarchies (caches, DRAM), and interconnects—especially PCIe.
• Experience reading performance counters and implementing tracing via drivers or user-space services.
• Familiarity with profiling and analysis tools (e.g., Nsight, uProf, VTune, perf, Tracy, ETW).
• Ability to design analysis passes that detect memory inefficiencies and bandwidth saturation.
• Comfort working closely with hardware, firmware, driver, compiler, and ML application engineers.

Nice-to-have qualifications:

• Proficiency in Python for data processing and tooling scripts.
• Experience developing performance analysis or debugging tools and kernel-mode driver development (Linux or Windows).
• Familiarity with ML accelerator architectures, multi-host accelerator systems, firmware/embedded development, or HDL/verification.
• In-depth knowledge of PCIe protocol and analyzers; understanding of compiler internals and code generation.

Education Requirements

Not specified.

About the Company

Company: Etched

Headquarters: San Jose, CA, United States

Etched develops purpose-built AI inference ASICs and systems optimized for transformer models, aiming to deliver significantly higher performance, lower cost, and lower latency than GPUs. The company focuses on enabling applications like real-time video generation and advanced reasoning agents, and is backed by leading investors and engineers.