AI Infrastructure Software Engineer

Responsibilities & Objectives Participate in building the AI Infrastructure technical architecture for JD.com's cross-border e-commerce; develop and drive the implementation of architectural standards. Lead the top-level architectural design for complex technical projects; evaluate critical technical issues and break down project execution steps. Scheduling System Optimisation: Participate in the R&D of the intelligent computing operating system; Optimize GPU resource scheduling capabilities within Kubernetes or Ray. High Availability Assurance: Resolve GPU cluster fault recovery (Fault Tolerance) and elastic fault tolerance (elastic checkpointing). Efficiency & Observability: Build cluster efficiency evaluation models; precisely measure computing resource utilisation (MFU/HFU). Track industry trends and cutting-edge technologies; introduce new technologies and solutions to the team based on actual business needs. Requirements: 3+ years of experience in AI infrastructure, distributed systems, High-Performance Computing (HPC), or large-scale cloud platform development. Proficient in Python, with a solid foundation in data structures and algorithms, and a rigorous coding style. Proficient in GPU/NPU architecture, with hands-on development and optimisation experience in CUDA C/C++ or Triton. Strong self-drive and technical curiosity; ability to proactively track the most cutting-edge AI Infra trends in the industry. Bachelor’s degree or higher in Computer Science, Electronic Information, Automation, Applied Mathematics, or related fields; experience in cross-border e-commerce is highly preferred. Deep understanding of the underlying source code and runtime mechanisms of PyTorch, Megatron-LM, or DeepSpeed. Fluent in both Chinese and English.