Etched.ai

Kernel Driver Software Engineer

San Jose$150k–$275kfulltimemidAdded today

About this role

Etched is seeking a Kernel Driver Software Engineer to design and optimize kernel-mode drivers for custom AI inference accelerators. You'll collaborate with hardware engineers on ASIC co-design, implement advanced virtualization and memory management, and ensure high performance and security for demanding machine learning workloads.

What you'll do

Design and maintain kernel-mode drivers with focus on reliability, debugging, and performance optimization
Optimize driver performance for AI workloads, minimizing latency and maximizing throughput
Implement virtualization support including SR-IOV, VFIO, and para-virtualization technologies
Design efficient memory management with kernel mapping, page tables, NUMA awareness, and IOMMU configuration
Build security features for host isolation, memory protection, and device attestation
Diagnose complex driver issues using kernel debugging tools like ftrace and kgdb

What they're looking for

C/C++ proficiency
Kernel-mode driver development and debugging
Linux operating system internals
PCIe and hardware interface protocols
Memory management and synchronization in kernel environments
Device virtualization technologies (SR-IOV, VFIO)
Kernel memory mapping, page tables, and IOMMU configuration
Hardware-software co-design principles

Likely interview questions

Walk us through your experience developing kernel-mode drivers. What was the most complex driver issue you debugged, and what tools did you use?
Describe your experience with memory management in kernel environments—specifically page tables, IOMMU configuration, and NUMA awareness. How have you optimized these for high-bandwidth data transfers?

Tell us about your experience with PCIe and device virtualization technologies like SR-IOV and VFIO. How would you approach integrating these into a new accelerator driver?
How would you design synchronization mechanisms to handle concurrent access to multiple accelerators, and what trade-offs would you consider for latency-sensitive AI inference workloads?
Describe your approach to performance profiling and optimization of kernel drivers. What metrics would you prioritize for an ML accelerator focused on inference throughput and latency?
Have you collaborated with hardware engineers during ASIC design or bring-up? How would you balance driver design decisions with hardware constraints?
What debugging tools and techniques would you use to diagnose a performance bottleneck in a driver handling ML workloads, and how would you validate your fix?

Unlock all 7 questions free — and practice them live →

Apply on the employer's site →

Opens the official application on the employer’s site. No login required.