Low-Level Kernel & CUDA Engineer

Company Description

Itur AI is building a proprietary, hardware-native acceleration stack that skips the standard runtime fluff for deterministic AI performance. You’ll be engineering the C++ driver layer that serves as our core technical moat.

The Mission

Driver Playground: Build a custom C++ driver for direct hardware and CUDA orchestration.

Role Description

This is a full-time and fully remote role for a Low-Level Kernel & CUDA Engineer. The role involves designing, developing, and optimizing low-level systems software, with a focus on CUDA and kernel development. The engineer will work on troubleshooting system-level performance issues, creating hardware-efficient algorithms, and collaborating with the team to enhance software architecture. You will also help define best practices and ensure functionality across a variety of hardware platforms.

Qualifications

• Mastery of C++ and CUDA (NVIDIA and AMD/ROCm) programming and GPU architecture

• Experience in low-level kernel development and system programming

• Serious experience with memory-mapped I/O and low-level system architecture

• Solid grasp of Transformer/MoE structures and what it takes to deploy at an enterprise grade

• Understanding of operating system concepts and hardware architectures

• Ability to debug and troubleshoot complex system-level issues

• Experience in multi-threading, parallel processing, and memory management

• Good communication skills and ability to work collaboratively in a remote environment

• Familiarity with AI/ML workflows, tensor libraries, or similar technologies is a plus