Member-only story
AI Engineer Interview Preparation #
Click** ** here for the full AI Engineer Prep list. 1. A large e-commerce company has an AI agent that processes millions of product images overnight. After processing 800,000 images, a worker crashes unexpectedly. Which architecture best minimizes repeated work after recovery?
(A) Save the entire application memory before shutdown
(B) Store progress after each completed batch in durable storage
(C ) Restart from the beginning to avoid inconsistent results
(D) Keep all progress only inside the worker’s memory
Correct Answer: (B) Store progress after each completed batch in durable storage.
An AI system running for hours cannot assume its process will always stay alive. The best design is to save durable checkpoints after meaningful batches so another worker can resume from the latest successful state. This reduces repeated processing, lowers infrastructure costs, and improves reliability. Checkpoints should contain resumable state such as completed batches, pending work, progress, and important metadata instead of temporary memory. This approach is widely used in distributed systems because failures are expected rather than…