机器学习平台调度工程师（北京/深圳/上海/杭州）
Core
Lead global resource scheduling for a 10,000+ GPU cluster to ensure efficient and stable operation of offline and online tasks.
Role type
Senior IC machine-learning platform scheduler (GPU cluster)
Builds
High-availability scheduling frameworks supporting distributed training frameworks
Domain
Cloud-native infrastructure + AI training infrastructure
Deliverable
production ML models
Required skills
Go/Python/C++, data structures and algorithms, Kubernetes core components, container runtimes, CRD development, OpenMP/MPI/RDMA high-performance computing, parallel computing principles, PyTorch/TensorFlow, operator optimization, memory management
Preferred skills
Hybrid cloud, virtualization, ARM heterogeneous computing
Technologies
Kubernetes, Docker, RDMA, CSI, CRD, PyTorch, TensorFlow
Responsibilities
Lead global resource scheduling for 10,000+ GPU clusters; Optimize RDMA high-speed networks and distributed storage/compute coordination; Build high-availability scheduling frameworks based on cloud-native technologies; Develop K8s schedulers, CSI plugins, and CRDs; Explore hybrid cloud, virtualization, and ARM heterogeneous computing.
