Determine the Requirements Needed for Inferencing CERTIFIED-AI-EXPERT Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A video-streaming platform uses a content-moderation model to flag inappropriate frames. The model processes 10,000 frames per second during peak hours, and each frame must be processed within 50 ms. The team is evaluating deployment options. Which configuration best meets the throughput and latency requirements?

   • AA CPU-only cluster with many cores
   • BA single high-end GPU server with a large batch size
   • CA cluster of GPU servers with load balancing and horizontal scaling
   • DEdge devices deployed at each content source
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which of the following is a common strategy to handle a sudden spike in inference requests?

   • AIncreasing the training dataset size
   • BDecreasing the model's learning rate
   • CAdding more inference server instances
   • DReducing the model's input resolution
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A team is deploying a real-time object-detection model on a fleet of edge devices with limited memory and compute. The model currently runs at 5 FPS, but the application requires at least 15 FPS. The team has already tried increasing the batch size, but the edge device cannot handle larger batches. Which optimization should they apply next?

   • AIncrease the number of edge devices and distribute the video frames among them.
   • BSwitch to a larger, more accurate model that runs at higher FPS.
   • CDeploy the model to a cloud GPU server and stream video to it for inference.
   • DApply quantization to reduce the model's precision from FP32 to INT8.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A robotics company is deploying an object-detection model on autonomous forklifts in a warehouse. The model must run in real time with a latency under 30 ms, and the forklifts have limited battery and compute resources. The team has a trained model that is 200 MB and requires 1 GB of RAM. Which set of requirements should they prioritize?

   • AHigh GPU compute and large memory to run the model at full precision.
   • BHigh throughput and large batch processing.
   • CCloud connectivity for offloading inference.
   • DLow latency, small model size, and low power consumption.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   Which of the following is an example of a software requirement for an AI inference workload?

   • AThe model must run on a GPU with at least 16 GB of memory
   • BThe network must provide a 10 Gbps connection
   • CThe data center must maintain a temperature below 25°C
   • DThe inference server must support the ONNX Runtime
   Reveal Solution Discussion Select an answer first