Multi-technology collaborative innovation: Liquid cooling technology must overcome challenges in efficient heat transfer and system reliability. Water-saving cooling sources need to achieve low water-efficiency temperature (WET) and coordinate dry and wet cooling methods. Power supply, distribution, and energy storage must solve source-grid-load-storage matching issues and enhance full-process efficiency. Coordinating innovations across these areas is highly complex.
Intelligent PUE optimization closed loop: Intelligent control of HVAC and power systems should be based on time sequence prediction and reinforcement learning to surpass human expert-level optimization. This demands advanced algorithms and strong engineering implementation.
Technology implementation and adaptation: Hardware innovation must comply with policies and fit industry needs, while AI operations must integrate with existing platforms and tools. Successfully combining and implementing these poses significant challenges....
Bridge the gap between cutting-edge Generative AI model research and real-world industrial applications by solving the “last mile” of AI deployment, ensuring models are performant, grounded, and effectively integrated into complex agent solutions that deliver direct business impact.
Design and prototype AI-native agent solutions based on client requirements, including agent workflows, system integrations, and LoRA fine-tuned models tailored to specific industry use cases, such as automated credit risk summaries for fintech or style-specific image and video generation.
Deliver production-ready implementations by writing high-quality code to integrate AI model APIs into agent frameworks such as ADK and LangChain, together with supporting components such as RAG, vector databases, memory systems, cache management, and skills....
Distributed Orchestration & Scheduling by providing support on distributed deployment of services/model training and inference tasks)
Optimize allocation rate and resource operation efficiency globally and coordinate capacity of various heterogeneous resources (CPU/GPU/heterogeneous hardware/training data)
Participate in development of training-related requirements (incremental training and API for online learning) and enhance framework efficiency...