AI-Ready GPU Server Architecture

I'm looking for an expert to design a GPU server architecture tailored for processing multiple computer vision AI models simultaneously. We need a robust and scalable solution to handle over 5 models at once.

Key Requirements:
- High-performance GPU selection
- Efficient data throughput and storage solutions
- Scalability for future model integration
- Detailed architecture design with implementation guidelines

Ideal Skills and Experience:
- Proven expertise in GPU server architecture
- Strong background in AI, particularly computer vision
- Experience with high-performance computing and data processing
- Knowledge of scalability and infrastructure optimization

Please provide previous work examples and a detailed approach.