AIF-C01 Fundamentals of AI and ML Practice Question

A financial services firm must build a model that flags potentially fraudulent card transactions in under 200 milliseconds while keeping all data inside its own Amazon VPC. The fraud team has thousands of labeled historical transactions and the pattern changes slowly over months. Which approach best balances latency, data residency, and the need for periodic retraining?

⚠ Common exam trap

The trap here is treating a managed fraud service or foundation model as automatically better than a purpose-built supervised model that meets the stated latency and residency constraints.

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

✓

Train a supervised classification model with Amazon SageMaker, deploy it to a real-time endpoint inside the VPC, and schedule periodic retraining jobs.

A supervised classification model trained on the labeled history and deployed to a SageMaker real-time endpoint inside the VPC meets the latency and residency constraints, while scheduled retraining handles gradual fraud pattern drift. Purpose-built supervised learning uses the available labels, and in-VPC endpoints keep transaction data within the controlled network. Batch, unsupervised, or general foundation-model approaches each miss at least one hard requirement.

Answer analysis

Option-by-option breakdown

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

✓
Train a supervised classification model with Amazon SageMaker, deploy it to a real-time endpoint inside the VPC, and schedule periodic retraining jobs.
Why this is correct
Fraud flagging with thousands of labeled transactions is a supervised classification problem, and a SageMaker real-time endpoint keeps inference within the VPC at low millisecond latency. Scheduled retraining jobs let the model adapt as fraud patterns drift over months. This combination satisfies the latency, residency, and retraining requirements directly without introducing external data movement.
✗
Deploy a pre-trained foundation model from Amazon Bedrock and prompt it to classify each transaction.
Why it's wrong here
Foundation models in Amazon Bedrock are general-purpose and not trained on the firm's labeled transaction history, so accuracy on a specialized fraud task would be unreliable. Prompt-based classification also typically has higher and more variable latency than a purpose-built model endpoint. This approach neither uses the available labels effectively nor guarantees the sub-200-millisecond budget.
✗
Build an unsupervised anomaly detection model on unlabeled data and run batch transform once per day.
Why it's wrong here
Batch transform runs on a schedule and returns results after the job completes, so it cannot deliver sub-200-millisecond decisions at transaction time. Discarding the thousands of labeled examples also wastes the most valuable signal for fraud detection. Unsupervised anomaly detection is useful when labels are absent, but here labels exist and real-time scoring is required.
✗
Use Amazon Fraud Detector in evaluation mode and export predictions to an S3 bucket outside the VPC for scoring.
Why it's wrong here
Exporting predictions outside the VPC violates the data residency requirement, and evaluation mode does not produce production scoring decisions. Amazon Fraud Detector itself is a managed service that can be used within a VPC via interface endpoints, but the described export workflow would move transaction data outside the controlled network boundary. This option fails the residency constraint and does not meet the low-latency production scoring need.

Go deeper

AIF-C01 Fundamentals of AI and ML Practice Question

Why each option matters

Option-by-option breakdown

Related to this question