Classify millions of product descriptions overnight.
No per-item instant response is required, but results are needed by morning.
Choose two appropriate designs.
Use batch inference with managed input and output units, and monitor completion time, retries, cost, and quality against the deadline.
Detailed explanation
The design can be checked against the morning deadline.
Rate limits, cost, and failure recovery become difficult.
Large workloads can be grouped without a real-time connection for every item.
Batch output still needs quality and safety checks.
Check official documentation
Compare the following points with the official documentation.
Amazon Bedrock公式推論方式とAIF-C01 Domain 3.1のモデル選択を確認する。
Official references
What to confirm
リアルタイム性と処理量・運用費用の要件で推論方式を選べる。
Key points
- Batch inference
- SLO
- Retry
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
Batch inference
Processing a collection of inputs as a job and retrieving results after processing.
Job manifest
A record identifying the input items, output location, settings, and processing state of a job.