资深异构计算架构师（J84309）
Core
Design and build an industry-leading AI heterogeneous computing container platform for high-performance, stable, and easy-to-use hybrid cloud products, supporting efficient deployment of AI applications in AIGC, intelligent computing centers, autonomous driving, and financial/energy sectors.
Role type
Senior IC AI Heterogeneous Computing Architect
Builds
Full-stack AI application cloud-native solutions, heterogeneous resource scheduling systems, and systematic acceleration schemes for model training and inference.
Domain
Cloud-native AI, Heterogeneous Computing, Large Model Training & Inference
Deliverable
production ML models
Required skills
Kubernetes architecture and development, Deep learning frameworks (PyTorch, TensorFlow, PaddlePaddle), GPU chip architecture, Distributed training optimization, Inference optimization (load balancing, model compression, caching), Heterogeneous computing acceleration, Container runtime and networking, Golang/Python/Java/C/C++ programming, Algorithm and data structures.
Preferred skills
Experience with large model training/inference frameworks (Megatron, vLLM), Virtualization and mixed deployment, High-performance networking, MoE architecture tuning, Zero/Offload memory optimization.
Responsibilities
Architect cloud-native AI container products, integrate open-source AI frameworks and scheduling components, engineer SOTA model optimization principles into production, explore innovations in task management and distributed training for large-scale clusters, participate in open-source communities to enhance technical influence.
Seniority
Senior, hands-on IC
