Free AWS AIP-C01 Actual Exam Questions - Question 3 Discussion

Good point on synonyms, but runtime hints (B) seem faster without structural changes. B

B Adding runtime hints can quickly steer Lex to recognize those variations without changing the slot type or intents, fitting the need for an immediate fix with minimal effort.

C imo. Defining synonyms in the slot type is a neat way to handle variations like “thrill-seeking” mapping to “adventure” without touching intents or Lambda. It keeps recognition tight and straightforward. Adding new enumeration values (D) works too, but synonyms let you cover multiple related terms without bloating the list. Runtime hints (B) are handy, but might not be as reliable or persistent for all cases. A (new intents) feels like overkill here since it complicates the model unnecessarily.

B/D? Adding runtime hints (B) is a quick way to guide Lex without structural changes, perfect for immediate fixes. But updating the slot type list (D) also ensures those words are officially recognized, which feels more reliable long-term.

Maybe D makes the most sense here since updating the slot type to include those new words means Lex can directly match them without any extra guesswork at runtime. It keeps things simple and avoids adding new intents or complicated logic. Plus, it doesn't require backend changes like the Lambda or database.

I’m skipping A because creating a whole new intent for every variation sounds like overkill. C is tempting but if the slot type doesn’t support synonyms natively, then D’s just a safer, more straightforward fix.

B. Adding runtime hints can quickly nudge Lex to recognize similar inputs without changing slot definitions or intents, making it a fast fix for variations like "thrill-seeking."

C, since synonyms directly map variations without adding new intents or complex updates.

C/D? Adding synonyms (C) can directly link those variations without changing logic, which seems quick and clean. D would help too by including new words in the slot type, but that might require more validation to avoid conflicts. A feels like overkill since creating a new intent is heavier and might confuse the bot. B might help slightly, but runtime hints don't always catch less obvious synonyms like “thrill-seeking.” So between C and D, I’d pick C for immediate recognition improvements without structural changes.

I get why B looks simple, but loading the model at runtime from S3 might add latency, which isn’t great for real-time inference. Option A uses Lambda to trigger SageMaker endpoints — that could add a bit of overhead but keeps inference tightly managed and secure. D seems to cut down on infrastructure by importing directly, which might actually reduce overhead more than spinning up Lambdas or EC2. But I’m also unsure if all fine-tuned SageMaker models are compatible with Bedrock’s import feature. Has anyone confirmed if D supports all model types or has any hidden limitations on real-time monit

B imo, because storing in S3 and letting Bedrock load at runtime avoids extra infrastructure like Lambda or EC2. Plus, CloudWatch monitoring fits naturally here without adding ops complexity.

Option D sounds promising for minimizing overhead, but does Bedrock AgentCore support direct imports of all SageMaker fine-tuned models? Also, is there any info on latency or compatibility limits?