Principal Architect – Hardware Efficient AI Foundation Model Training
Core
Lead the design of foundational model architecture for LLM/Code/Multimodal subfields by breakthroughs in post-training and continual training to enhance algorithm performance and training efficiency.
Role type
Principal Architect, Hardware Efficient AI Foundation Model Training
Builds
Foundational models with state-of-the-art performance and hardware efficiency for LLM, Code, and Multimodal applications
Domain
AI Systems, Chip Design, Distributed Training Infrastructure
Deliverable
production ML models
Required skills
Large-scale distributed training and inference, AI model training and optimization (10B+ parameters), AI architecture design (long-sequence, reinforcement learning, multimodal, agents), AI framework implementation (PyTorch, vLLM, SGLang), GPU/NPU/TPU architecture understanding, memory hierarchy and interconnect technologies
Preferred skills
PhD in AI architecture or computer architecture, publication records in AI systems or chip design
Technologies
PyTorch, vLLM, SGLang, GPU, NPU, TPU
Responsibilities
Collaborate with internal and external organizations to lead the design of foundational model architecture, Propose technical requirements for large-scale distributed training and inference infrastructures such as parallelization and operator fusion, Analyze the computational characteristics of typical architectures
Seniority
Principal, strategy & mentorship