Linux Device Driver Engineer (AI)

TL;DR

Linux Device Driver Engineer (NPU): Developing and maintaining Linux device drivers for NPUs to support high-performance AI computation with an accent on PCIe subsystems, DMA operations, and memory management. Focus on optimizing data transfer latency, implementing safe memory abstractions using Rust-for-Linux, and debugging complex system-level hardware-software interfaces.

Location: Onsite in Seongnam, South Korea

Company

AI chip startup specializing in the development of high-performance NPUs.

What you will do

• Develop and maintain Linux device drivers for NPUs to enable high-performance AI computation.
• Design high-throughput, low-latency data transfer solutions and complex DMA operations over PCIe subsystems.
• Architect and optimize memory management subsystems to ensure stability and zero-overhead host-device communication.
• Investigate and adopt modern kernel paradigms, including Rust-for-Linux, to eliminate memory reference vulnerabilities.
• Design and implement communication interfaces with device firmware for seamless hardware integration.
• Support and debug complex system-level issues across the software stack and hardware interfaces.

Requirements

• Bachelor’s or higher degree in Computer Science, Electrical Engineering, or related fields.
• Minimum of 5 years of experience in general Linux kernel and device driver development (PCIe, DMA, Network, Storage, DRM).
• Deep understanding of SR-IOV, PCIe device virtualization, IOMMU/SMMU, and DMA (dma-mapping, dma-buf, scatter-gather).
• Strong command of kernel concurrency (locking, RCU, workqueues) and ability to debug deadlocks and use-after-free.
• Proficiency in C, C++, and Rust.
• Proven experience in low-level debugging using kernel logs, KASAN/ASAN, and coredumps.

Nice to have

• Hands-on experience with Rust for Linux, specifically zerocopy or safe abstractions for device drivers.
• Previous experience in GPU, NPU, or other complex accelerator driver development (DRM subsystem).
• Understanding of device memory communication protocols and zero-copy data movement between user-space and kernel/hardware.
• Over 5 years of experience in performance tuning for large-scale systems.

Hiring process

• Document screening.
• Online interview, including a coding test.
• On-site interview, including a practical assignment.
• Compensation negotiation and final offer.