REQUIREMENTS

To ensure your success in this role, the following experience is needed.
Must have:
• 2+ years of hands-on experience with Yocto/OpenEmbedded (BitBake, layers, classes, packagegroups, etc.).
• Strong C (kernel and native), plus shell and Python; comfortable with C++ runtime libraries.
• Experience with AI/ML inference runtimes and accelerators (NPU/GPU).
• Strong understanding of multi-core application SoCs: CPU clusters, interconnect/NoC fabric, memory map and address decoding, clocks/resets/power domains, and cache coherency.
• Hands-on accelerator and peripheral bring-up and debugging, with proven experience debugging the full host-to-accelerator data path (NPU/GPU/DSP).
• Hardware and board debugging: serial console, dmesg, gdb, and Lauterbach

Joining our team will have the opportunity to:
• Work on AI platforms that ship in hundreds of millions of devices — automotive, industrial, IoT, and consumer
• Collaborate with world-class silicon architects, compiler engineers, and AI researchers
• Shape the Edge AI ecosystem at a company with 50+ years of semiconductor leadership
• Influence open-source inference frameworks and industry standards from a position of silicon authority

JOB DESCRIPTION

Your responsibilities

• BSP integration of the AI/ML runtime — integrate the accelerator runtime and inference libraries into the embedded Linux image.
• Yocto/OpenEmbedded recipe development and maintenance — author and maintain recipes, packagegroups, and layer configuration.
• Cross-team and vendor coordination — drive integration across the kernel/platform, silicon/BSP, runtime, and application teams; collaborate with silicon and runtime vendors, track release cadences, and file precise, reproducible defect reports that drive upstream fixes.
• Cross-functional debugging — when an issue spans layers (e.g., a fault that could originate in a recipe, driver, or silicon), own the investigation, isolate the responsible layer, and hand each team an actionable, evidence-backed writeup.
• Cross-platform enablement — unify the runtime across multiple SoC/BSP targets, minimizing per-platform forks in favor of a shared, modular layer.