锦鲤求职岗位描述
About Kiwimoore:
Kiwimoore is at the forefront of the semiconductor industry, pushing the boundaries with our innovative Chiplet technology. We are dedicated to advancing computing through a culture of collaboration, innovation, and dynamic teamwork. As part of Kiwimoore, you will join a team committed to redefining the future of semiconductor technology with some of the brightest minds in the industry.
Role Overview:
Kiwimoore is seeking a Principal AI Networking Software Architect to define the long-term technical vision and end-to-end software architecture for next-generation AI networking solutions. The role will translate the communication requirements of large-scale AI training and inference workloads into scalable Scale-up and Scale-out network architectures spanning GPU/XPU systems, NICs, DPUs, switches, memory and storage. The architect will lead technology pathfinding, hardware-software co-design, customer co-development, proof-of-concept validation, and cross-functional execution from research concept to production.
Location: Shanghai,Beijing
Responsibilities:
•Set the long-term technical vision, architecture and multi-generation roadmap for distributed AI communication, including GPU-to-GPU, GPU-to-storage and cross-node data movement.
•Characterize the communication requirements and bottlenecks of large-scale AI training and inference workloads, including collective communication, model parallelism, KV cache transfer and disaggregated prefill/decode.
•Define cohesive end-to-end Scale-up and Scale-out architectures across GPU/XPU compute, NICs, DPUs, Ethernet switches, memory systems, optical interconnects and software stacks.
•Research and prototype next-generation networking solutions using RDMA/RoCE, Ultra Ethernet, GPUDirect, PCIe/CXL and emerging accelerator interconnect technologies.
•Drive hardware-software co-optimization across transport, communication libraries, host networking, firmware, drivers, SDKs and programmable network devices.
•Work across architecture, switching, NIC, optics, silicon, firmware and software teams to align technical strategy, interfaces, feature priorities and product roadmaps.
•Engage hyperscalers, CSPs, AI infrastructure providers, GPU/XPU vendors, OEMs and ODMs in deep co-design discussions covering cluster architecture, topology, congestion control, transport behavior and reliability.
•Translate customer and workload requirements into actionable architecture specifications, product requirements, reference designs and validation plans.
•Build proof-of-concept systems, simulations and quantitative performance models; validate architecture choices on representative AI workloads and clusters.
•Integrate networking capabilities with AI communication and serving stacks such as NCCL, UCX, MPI, NVSHMEM, NIXL, vLLM, SGLang or TensorRT-LLM.
•Identify portfolio gaps and incubate new protocols, software features and product concepts while ensuring interoperability and competitive differentiation.
•Represent Kiwimoore in relevant standards bodies and industry forums, and develop architecture specifications, white papers, technical presentations, patents and customer-facing documentation.
•Provide technical leadership and mentorship to senior engineers and influence complex cross-team initiatives from research through production.
Qualifications:
Minimum Qualifications:
•Master's or Ph.D. degree in Computer Science, Computer Engineering, Electrical Engineering, Communications, Microelectronics or a related field, or equivalent practical experience.
•12+ years of experience in systems software, networking architecture, Ethernet switching, NIC/DPU development, distributed systems or HPC/AI infrastructure, with a record of delivering complex cross-functional initiatives.
•Deep expertise in high-performance networking, including Ethernet, InfiniBand, RDMA/RoCE, transport protocols, congestion control, load balancing, routing and performance optimization.
•Strong understanding of AI/ML training and inference workloads, GPU/XPU communication patterns, collective communication and distributed execution models.
•Strong understanding of computer architecture, memory hierarchies, DMA engines, PCIe/CXL, HBM, NIC/DPU/switch architecture and operating-system networking.
•Hands-on experience with communication libraries or runtimes such as NCCL, UCX, MPI, NVSHMEM, SHMEM or NIXL.
•Proficiency in C/C++ and Python; ability to prototype, debug and analyze performance across hardware and software boundaries.
•Demonstrated ability to translate ambiguous workload and customer requirements into architecture, specifications, roadmaps and measurable performance targets.
•Proven ability to influence product strategy and technical direction across organizations, with excellent written and verbal communication skills in English.
Preferred Qualifications:
•15+ years of relevant experience and prior ownership of a company-level or multi-product AI networking architecture.
•Experience with Ultra Ethernet/UEC, UALink, high-radix switching, Clos or Dragonfly topologies, multi-plane Scale-up fabrics and custom AI transports.
•Experience with AI inference frameworks and communication requirements, including vLLM, SGLang, TensorRT-LLM, disaggregated serving and KV cache movement.
•Experience with GPU programming, GPUDirect, CUDA, accelerator architecture or topology-aware collective algorithms.
•Experience with optical interconnects and system implications of pluggable optics, LPO or CPO.
•Experience working with hyperscalers, major CSPs, AI model companies, GPU/XPU vendors and system OEM/ODM partners.
•Active participation in standards or industry bodies such as UEC, OCP or IEEE.
•Track record of patents, publications, major open-source contributions or presentations at leading technical conferences.
What Kiwimoore Offers:
1.A competitive salary and comprehensive benefits package.
2.Career growth opportunities in an innovative and dynamic field.
3.A creative, inclusive, and collaborative work environment.
4.The opportunity to be part of a pioneering team revolutionizing technology.
How to Apply:
If you are a dynamic professional looking to impact the tech world , apply now! Submit your application through our careers page, including a resume and a cover letter explaining why you are the perfect fit for the Fellow, Systems Architecture at Kiwimoore.
Kiwimoore is an equal opportunity employer. We value diversity and are committed to creating an inclusive environment for all employees. All employment is decided based on qualifications, merit, and business need.