AI Frameworks Engineer – GPU Performance for Generative AI (OpenVINO)

Job Details: Job Description: Role Overview We are seeking a software engineer to drive the implementation and performance optimization of generative AI workloads on Intel GPUs as part of the OpenVINO GPU team. This role focuses on building high-performance, HW-aware software that enables efficient execution of AI models on current and future Intel GPU architectures. You will work across multiple layers of the stack—AI models, runtime systems, and GPU hardware—and take ownership of complex performance problems that require deep technical insight and careful trade-off analysis. You will work on state-of-the-art AI models that push the limits of GPU performance. Your work directly impacts real-world AI performance experienced by developers and customers. About OpenVINO OpenVINO(https://github.com/openvinotoolkit/openvino) is a performance-focused AI inference runtime designed to efficiently execute deep learning models across Intel architectures. The GPU plugin is a core component of OpenVINO that bridges high-level AI models and low-level GPU execution, covering areas such as graph transformation, kernel dispatch, memory management, and hardware-specific optimizations. The codebase is performance-critical, largely written in modern C, and requires strong understanding of system-level software design, debugging, and optimization. What You Will Do Take technical ownership of performance-critical paths for generative AI workloads (e.g., LLMs, diffusion models) on Intel GPUs Analyze end-to-end execution of AI models to identify compute, memory, bandwidth, and parallelism bottlenecks Implement and optimize generative AI techniques, adapting state-of-the-art ideas to efficiently run on Intel GPU architectures Translate deep understanding of GPU hardware architecture into efficient, scalable, and maintainable software designs Optimize workloads for both current and future Intel GPU platforms, including hardware that is still under development Diagnose and resolve complex issues that span runtime, kernel, driver, and hardware boundaries Collaborate with global teams across software, hardware architecture, and validation to deliver optimized solutions Qualifications: Required Qualifications Computer science, computer engineering, or a related field with 3 years of professional software engineering experience Strong programming skills in C and C; working experience with Python Experience working with large and complex C codebases, with attention to performance, correctness, and maintainability Proven analytical thinking and strong problem-solving abilities, especially for ambiguous or under-specified problems Preferred Qualifications Experience with GPU programming or parallel computing, such as multi-threading, SIMD, or accelerator programming models Strong understanding of computer and GPU architecture, and how hardware characteristics impact software performance Technical understanding of generative AI models from a system and performance perspective Familiarity with AI runtimes or frameworks Solid foundation in computer science fundamentals, including data structures, algorithms, and operating systems Ability to communicate technical ideas clearly in written and spoken English Work Model This role follows a structured hybrid work model. The team regularly combines remote work and in-office collaboration, with a designated in-office days each week, while the remaining days are remote. Job Type: Experienced Hire Shift: Shift 1 (Korea, Republic of) Primary Location: South Korea, Seoul Additional Locations: Business group: The Software Team drives customer value by enabling differentiated experiences through leadership AI technologies and foundational software stacks, products, and services. The group is responsible for developing the holistic strategy for client and data center software in collaboration with OSVs, ISVs, developers, partners and OEMs. The group delivers specialized NPU IP to enable t ... (truncated, view full listing at source)