Câu 18: NCP-AAI: Agentic AI

An e-commerce platform is implementing an AI-powered customer support system that handles inquiries ranging from simple FAQ responses to complex product recommendations and technical troubleshooting. The system experiences unpredictable traffic patterns with sudden spikes during sales events and varying complexity requirements. Simple questions comprise the majority of requests but require minimal compute, while complex product recommendations need sophisticated reasoning. The company wants to optimize costs while maintaining service quality across all query types. Which approach would provide the MOST cost-optimized scaling strategy for this variable-workload, mixed-complexity environment?

1. A. Deploy specialized NVIDIA NIM microservices using a single large model configuration that handles all agent functions on high-capacity GPUs, with auto-scaling infrastructure that maintains constant resource allocation across all traffic patterns.
2. B. Deploy specialized NVIDIA NIM microservices on CPU-optimized infrastructure with auto-scaling capabilities to minimize hardware costs, while accepting longer inference times for cost optimization benefits.
3. C. Deploy specialized NVIDIA NIM microservices with an LLM router to dynamically route requests to appropriate models based on complexity, combined with auto-scaling infrastructure that scales different model types independently. — đáp án hiện tại
4. D. Deploy multiple specialized NVIDIA NIM microservices with identical high-capacity models across all available GPUs, implementing auto-scaling infrastructure without request complexity differentiation or dynamic model selection capabilities.