AIF-C01 Applications of Foundation Models Practice Question

A multinational corporation uses a foundation model via Amazon Bedrock to translate internal communication documents from English to multiple languages. They notice that the translations often miss company-specific jargon and acronyms, leading to confusion. The company has a glossary of approved translations for terms like 'Project Atlas' and 'Operation Synergy.' They want to improve translation accuracy quickly and with minimal effort. What approach should they take?

⚠ Common exam trap

The trap here is that candidates may overestimate the effort required for prompt engineering or underestimate the speed and simplicity of in-context learning, leading them to choose fine-tuning or a different service when the most direct and minimal-effort solution is to augment the prompt with the glossary.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Use prompt engineering to include the glossary in each translation request.

Prompt engineering allows the company to inject the glossary directly into the context window of the foundation model with each translation request. This approach requires no model retraining or infrastructure changes, enabling rapid improvement by simply appending the approved translations as instructions or few-shot examples. It is the quickest and least effortful method to enforce company-specific terminology without altering the underlying model.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✓
Use prompt engineering to include the glossary in each translation request.
Why this is correct
Embedding the approved glossary directly in each prompt supplies the model with the exact company-specific terms and acronyms at inference time, requiring no training or fine-tuning. This satisfies the minimal-effort, quick-improvement constraint for jargon like 'Project Atlas'.
✗
Use a larger foundation model that has better language understanding.
Why it's wrong here
A larger model still lacks the company's glossary, so 'Project Atlas' and 'Operation Synergy' remain mistranslated regardless of general language capability. It is tempting because larger models do improve general translation fluency, and would be the right choice if the problem were poor grammar in common vocabulary rather than missing proprietary terms.
✗
Fine-tune the foundation model on a corpus of bilingual company documents.
Why it's wrong here
Fine-tuning retrains model weights on bilingual documents, a slow, costly process requiring curated parallel corpora and redeployment, contradicting the minimal-effort, quick requirement. It is tempting because fine-tuning genuinely adapts a model to domain-specific terminology, and would suit scenarios where no glossary exists and large volumes of labelled translation pairs are already available.
✗
Switch to Amazon Translate with custom terminology.
Why it's wrong here
Amazon Translate is a separate service, so switching abandons the existing Bedrock foundation-model workflow and requires re-integration rather than a quick configuration change. It is tempting because custom terminology is precisely designed to enforce glossary mappings, and would be correct if the workload were pure translation without any Bedrock dependency.

Go deeper

AIF-C01 Applications of Foundation Models Practice Question

Why each option matters

Option-by-option breakdown

Related to this question