Senior Software Engineer, AI Compilers

About the Team

This position is part of the Kinara Group at NXP Semiconductors, a team combining startup agility and innovation with the global scale and stability of a leading semiconductor company. Kinara originated from patented Stanford University Ph.D. research, focusing on revolutionizing AI inference at the edge through integrated hardware and software solutions. Now operating within NXP, the Kinara team develops advanced AI accelerators and software dedicated to edge AI, delivering proven silicon generations and a mature software stack used by top-tier global customers.

Kinara technologies power high-profile clients including major technology companies, a leading e-commerce giant, and Tier-1 PC manufacturers. With NXP's manufacturing capabilities and customer access, the team balances innovation, expertise in AI, and a compiler-driven engineering approach to create cutting-edge AI compiler and silicon technologies with measurable impact and stability.

Our Work

We design revolutionary edge AI platforms that enable faster, smarter, and more energy-efficient decisions where low latency is critical. Our core products include the Ara AI inference processors built to maximize deep learning performance per watt on edge devices. Paired with an extensive SDK and toolchain, these processors empower developers to deploy and optimize complex AI models for real-time inference under stringent latency and power limits.

Through a hardware-software co-design strategy, we embed high-performance AI into a wide variety of edge applications, such as vision systems, industrial automation, consumer electronics, and intelligent devices. Our focus on performance, efficiency, and scalability helps customers transition sophisticated AI capabilities from cloud environments directly into practical, real-world uses. Positioned at a pivotal moment for edge AI development, the Kinara Group within NXP is a leading force, delivering proven silicon, comprehensive compiler and SDK tools, and a roadmap for next-generation intelligent edge solutions.

Responsibilities

• Design and implement compiler components across subsystems including MLIR dialects, lowering passes, operator fusion, and target-specific code generation for NPU hardware.
• Develop quantization workflows and numerical correctness verification for assigned compiler features.
• Create and release robust, efficient, and maintainable compiler passes and runtime integration pieces suitable for production environments.
• Optimize low-level compiler performance and collaborate with kernel teams to resolve hardware/software performance bottlenecks.
• Contribute to continuous integration pipelines, automated test suites, profiling tools, and release and deployment processes for the compiler.
• Investigate performance and correctness issues in client models, providing reproducible solutions and fixes.
• Guide and mentor junior engineers, participating actively in design and code reviews to elevate team quality.
• Offer insights to product and silicon teams on compiler-driven enhancements and potential trade-offs impacting hardware and software features.

Required Qualifications

• 7 to 8 years of professional software engineering experience, especially in compiler systems development.
• Expertise in graph lowering, intermediate representation (IR) design, operator fusion techniques, scheduling algorithms, and target code generation tailored for accelerator architectures.
• Proficiency in system-level programming using C++ and strong Python skills for tooling and build infrastructure.
• Proven experience delivering compiler or system software components that are deployed in production products.
• Advanced debugging and performance profiling capabilities using industry-standard tooling.
• Excellent communication skills with the ability to collaborate effectively across hardware architecture and kernel development teams.
• A commitment to mentoring junior team members and enhancing engineering best practices within the group.

Preferred Experience

• Familiarity with MLIR frontend frameworks and dialects like TOSA, StableHLO, or TorchMLIR.
• Background in AI compiler stacks, backend hardware development, or targeting accelerator devices.
• Hands-on experience with large-scale machine learning models or production-level inference pipelines.