Software Engineer – Performance Profiling

About this role

Etched is seeking a Software Engineer to design and build a comprehensive performance analysis tool for their custom ML accelerator hardware. You'll create data collection mechanisms, tracing systems, and visualizations that help engineers and customers understand workload behavior and identify bottlenecks across hardware, drivers, and system-level events.

What you'll do

• Design and architect a complete performance analysis suite including data collection, processing pipelines, and user interfaces
• Develop methods to capture performance data from custom ML accelerator hardware via performance counters and driver interfaces
• Implement tracing for host-side API calls and system-level events across CPU, PCIe, and memory systems
• Build analysis modules to automatically identify performance bottlenecks and limiters from trace data
• Create intuitive visualizations like timelines and dependency graphs to communicate performance characteristics
• Collaborate with hardware architects, firmware engineers, and ML engineers to define tool requirements and provide optimization guidance

What they're looking for

• C++ or Rust proficiency
• Computer architecture knowledge (CPU, GPU, accelerators, memory hierarchies, PCIe)
• Low-level performance analysis and profiling on complex hardware systems
• Experience with performance analysis tools (VTune, Nsight, perf, Tracy, etc.)
• Device driver interaction and hardware performance counter reading
• Python (preferred)
• Data visualization and UI design
• ML accelerator architecture understanding (nice-to-have)

Likely interview questions

• Walk us through your experience building or working with performance analysis tools. What was the most complex bottleneck you identified and how did you help resolve it?
• Describe your experience capturing and interpreting hardware performance counters. What challenges have you faced in low-overhead data collection on complex systems?