NPU SDK Software Engineer

• To take care of the full life-cycle of our SDK features in one of the components, including the simulator and compiler, or the API and the infrastructure of our SDK;
• Set and uphold high standards for code quality, maintainability, testing, and release readiness;
• Conduct code reviews and improve engineering practices and ensure robust documentation within the expertise domain;
• Work closely with other team (Digital Design, Firmware, Applications), translate complex algorithm or hardware specifications into efficient, user friendly, and testable software behaviour, and provide technical feedback;
• Provide support to internal teams and customers to enable them to develop applications quickly using our SDK or help solve any issues;
• Contribute to improving power consumption performance, latency, and any performance bottlenecks for our efficient edge-AI solution;
• Keep innovating, bring latest trend of the technology and best practice to keep our product competitive in the market.

• Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, or a related field;
• Strong Python development skills;
• Extensive experience with PyTorch and its customization, including knowledge of internal mechanics;
• Proven track record of designing modular, maintainable and testable software architecture;
• Experience in low-power resource-constrained embedded programming;
• Familiarity with hardware/software co-design concepts;
• Highly flexible, with a readiness to step outside your immediate domain to drive team success;
• A strong growth mindset with an active drive to master new tools and technologies;
• Great communication skills.

Nice to have:

• Modern C++ development skills (including C++17/20) would be a big plus, or a willingness to expand the knowledge;
• Exposure to bare-metal firmware development;
• Solid understanding of computer architecture, particularly ARM or RISC-V-based systems, and basic digital design knowledge is a big plus;
• Familiarity with ML compilers like MLIR, IREE or PyTorch compiler (Dynamo);
• Familiarity with the runtime of TinyML or EdgeAI runtimes (TensorFlow Lite Micro, execuTorch);
• Knowledge of SNN (Spiking Neural Networks).

Whether you’re passionate about microarchitecture, excited by verification or physical design, driven by supporting operations, building strong teams, or mastering the details of accounting, we’d love to hear from you!