CCNA Foundation Model Applications Questions

Amazon Bedrock supports model evaluation jobs that can automatically score model outputs against reference data using metrics like BERTScore and ROUGE. This provides a repeatable, automated comparison across multiple foundation models, which matches the requirement. It removes manual scoring and produces consistent results that can guide model selection for the product description task.

On-demand mode bills per token or image, so cost scales directly with actual usage rather than a fixed provisioned commitment. This satisfies the stem's unpredictable traffic and cost-reduction constraint, since idle provisioned capacity would otherwise be paid for regardless of demand.

Amazon Comprehend Medical can identify protected health information (PHI) in text, enabling redaction before sending data to the model and before returning outputs to users. This helps comply with medical privacy regulations such as HIPAA. While Bedrock Guardrails also offer PII redaction, Comprehend Medical is specifically trained for medical entities. This action directly supports the privacy compliance requirement.

Amazon Kendra provides a fully managed, ML-powered enterprise search service that indexes internal documents and returns relevant passages, which Bedrock then feeds to the foundation model for grounded answers. It satisfies the RAG requirement without infrastructure management, unlike self-managed vector stores or OpenSearch clusters.

Knowledge Bases for Amazon Bedrock ingests documents from sources such as Amazon S3, chunks and embeds them into a vector store, and retrieves relevant passages at query time so the model grounds answers in those documents and can return citations. This directly satisfies the requirement to answer from the sales reports and avoid fabricated figures.

Including the entire conversation history in each prompt allows the model to see previous exchanges and maintain context. This is a common technique for multi-turn conversations with foundation models, as they are stateless and do not remember previous interactions. However, it increases token usage and may hit context length limits, so it should be managed carefully.

Amazon CodeGuru Security applies static analysis with detector rules tuned to identify security vulnerabilities and deviations from coding best practices, directly satisfying the requirement that generated code be checked for vulnerabilities. Unlike generic content filters, it inspects code semantics, catching issues such as injection flaws and insecure data handling within the Bedrock output.

Amazon Bedrock Guardrails allows you to define denied topics, which are subjects the model should avoid. By specifying 'personalized financial advice' as a denied topic, the guardrail will block the model from generating content on that topic. This is the appropriate feature to enforce regulatory compliance by preventing certain types of responses.

Retrieval-Augmented Generation grounds responses in authoritative source documents retrieved at inference time, so the model cites real product data instead of relying solely on parametric memory. This directly reduces fabricated or incorrect details in generated descriptions.

Amazon Transcribe performs automatic speech recognition to convert audio into text, and Amazon Bedrock provides access to foundation models that can summarize or rewrite that transcript into concise captions. This two-step pipeline matches the requirement: transcription followed by generative caption creation. Both are managed AWS services, so the company avoids building and operating its own speech and language models.

Fine-tuning on a balanced, representative dataset corrects skewed statistical associations in the training data, so the model's outputs reflect fairer distributions across demographic groups. This directly targets the root cause of bias rather than merely masking its symptoms at inference time.

Injecting retrieved passages into the prompt gives the model authoritative source text to condition on, so answers cite supplied evidence rather than relying on parametric memory. This grounds generation in the knowledge base, directly improving factual accuracy.

Chunking splits the long document into segments that fit the model's context window, letting each chunk be summarised without truncation. This satisfies the long-document constraint, since whole-document input would exceed context limits and lose detail, degrading summary quality.

Bedrock Model Evaluation runs automatic toxicity metrics against model outputs, directly satisfying the requirement to monitor generated text for toxic content. It provides scored, repeatable assessment rather than ad-hoc inspection, making it the most effective ongoing evaluation approach for this deployment.

Setting max_tokens caps the number of tokens the model generates, directly satisfying the requirement to limit response length. In Amazon Bedrock, this parameter bounds output size independently of the prompt, so the data scientist can truncate verbose answers without altering the input or switching models.

Distillation trains a compact student model to mimic the larger teacher, cutting inference compute and memory per token. Fewer parameters mean faster forward passes, directly addressing the real-time chat latency constraint rather than merely tuning prompts or batching.

Placing instructions at the end of the prompt, after the long context, leverages the model's tendency to pay more attention to recent tokens. Repeating key instructions after the article text reinforces them, improving adherence without changing the model or adding significant cost. This is a known prompt engineering technique for long-context scenarios.

SageMaker endpoints do not scale automatically unless an autoscaling policy is attached to the endpoint variant. Without Application Auto Scaling configured against a target metric such as InvocationsPerInstance, the endpoint keeps fixed instance count and cannot absorb rising load.

Amazon Bedrock Knowledge Bases performs managed retrieval-augmented generation by ingesting documents from Amazon S3 into a vector store, retrieving relevant passages at query time, and returning responses with source attribution. This directly satisfies the requirement for grounded, verifiable citations while reducing fabrication, because the model conditions its output on retrieved approved content rather than only on its pretrained parameters.

RAG grounds the model's response by supplying relevant, authoritative content from the source articles directly in the prompt. This reduces hallucination because the model conditions its output on the retrieved text rather than relying solely on parametric knowledge. It requires no retraining and integrates with Amazon Bedrock Knowledge Bases or custom retrieval, making it suitable for dynamic news content.

RAG with an Amazon Bedrock knowledge base indexes the policy documents and retrieves the most relevant passages for each query, then passes them to the model as context. This grounds responses in the latest authoritative content without retraining, and updating the knowledge base data source keeps answers current. It directly addresses outdated or incorrect policy details by supplying fresh source text at inference time.

Model access in Amazon Bedrock is granted separately from IAM permissions. Even with bedrock:InvokeModel allowed, the account must first request access to the Anthropic Claude model in the Bedrock console, otherwise invocation returns an access-denied error despite the permissive policy.

A smaller, distilled model reduces the number of parameters and floating-point operations per inference, directly cutting compute time on the SageMaker endpoint. This satisfies the stem's real-time latency constraint, since inference duration scales with model size; distillation preserves much of the original accuracy while delivering sub-second responses.

Amazon Bedrock provides serverless access to image generation foundation models, including Amazon Titan Image Generator, through the InvokeModel API. The startup sends a text prompt and receives generated images without provisioning GPUs or managing endpoints. This matches the requirement for a managed text-to-image API with no infrastructure ownership.

Content filters intercept prompts and completions at runtime, blocking harmful or biased output before it reaches users. This directly satisfies the stem's requirement to reduce risk from foundation models, since filtering operates on the model's generated content itself rather than on training data or access controls.

Supplying the specifications directly in the prompt grounds generation in authoritative source data, so the model conditions its output on those facts rather than relying on parametric knowledge that may be outdated or hallucinated. This directly satisfies the stem's constraint of reducing inaccurate specification details, since the model is instructed to base descriptions solely on the provided data.

Amazon Bedrock knowledge bases can return citations that link generated statements to the retrieved source chunks, and a system prompt can instruct the model to answer only when supporting context exists. This satisfies both requirements: source citations and abstention when no document is found. It leverages built-in retrieval and citation features rather than custom post-processing.

Amazon Bedrock offers access to foundation models that perform abstractive summarisation, condensing ticket text into a short paragraph. It provides a managed, serverless API without infrastructure management, making it the most appropriate service for this natural language generation task.

Amazon Bedrock natively emits metrics to CloudWatch and logs API activity to CloudTrail without custom instrumentation, satisfying the operational visibility requirement that self-managed EC2 inference stacks must build and maintain themselves. This removes undifferentiated monitoring plumbing, letting teams focus on model usage rather than telemetry infrastructure.

Amazon Bedrock knowledge bases provide a managed RAG workflow: they ingest and chunk source documents, generate embeddings, store them in a vector store, and retrieve relevant passages at query time. The model can then generate answers grounded in those retrieved passages. This removes the need for the developer to orchestrate chunking, embedding, and retrieval manually, matching the requirement for minimal application code.

Setting temperature to 0 makes the model select the highest-probability token at each step, eliminating the random sampling that produces run-to-run variation. Keeping top_p at 1 disables nucleus filtering, so it cannot reintroduce randomness. This satisfies the audit requirement for consistent, deterministic sentiment outputs from identical transcripts.

ThrottlingException indicates the account's requests per minute or tokens per minute exceed the model's allocated throughput. Retrying with exponential backoff and jitter spreads retries, letting transient capacity recover instead of compounding the overload with immediate repeat calls.

Guardrails for Amazon Bedrock includes sensitive information filters that can detect and block or mask personally identifiable information (PII) and custom regex patterns. By configuring these filters, the company can prevent PHI from appearing in model responses. This directly addresses the requirement to avoid including PHI in generated patient education materials.

Amazon Bedrock Knowledge Bases with a metadata filter satisfies the post-2023 constraint by restricting vector retrieval to documents whose publication-date metadata matches the filter, so the foundation model generates answers grounded solely in that retrieved subset rather than its training data. This enforces the temporal restriction at retrieval time, preventing older papers from entering the context window.

Amazon Bedrock offers a serverless, API-driven way to invoke foundation models from multiple providers, with usage billed per input and output token. That matches the stem's serverless API and per-token pricing constraints without managing infrastructure.

BLEU compares generated text against reference text using n-gram overlap, producing a precision-based score. It suits marketing copy evaluation where reference outputs exist. Perplexity, by contrast, measures language model likelihood without references, so it cannot assess similarity to target text.

Exponential backoff with jitter retries throttled requests after progressively longer, randomized delays, which spreads retries out and prevents synchronized retry storms. It is a small, well-understood code change that directly addresses transient ThrottlingException errors during bursts. AWS SDKs and the Bedrock runtime support configurable retry behaviour, making this the lowest-effort effective fix.

Summarisation is a text-generation task: the model consumes the source document as input and autoregressively produces a condensed natural-language output. This directly satisfies the stem's requirement to automatically summarise lengthy documents, since the capability being exercised is generating new text rather than classification, embedding or retrieval.

Action groups in Amazon Bedrock Agents use OpenAPI schemas to describe available operations. By defining an action group with the inventory query and reservation endpoints, the agent knows how and when to call them. This enables the model to perform the required actions through the agent's orchestration, making it a correct implementation step.

Selecting a domain-strong base model reduces the volume of task-specific examples and training steps required, since the model already encodes relevant vocabulary and structure. This directly satisfies the stem's optimisation goal by lowering compute cost and convergence time during Bedrock fine-tuning, rather than compensating for weak domain representation through extra data.

Enabling the safety checker activates Bedrock's built-in content filters, which evaluate each generated image against violence and inappropriate-content thresholds before returning it. This directly satisfies the agency's need to block harmful output automatically at generation time, rather than relying on manual review or post-hoc moderation of images already displayed.

Amazon ElastiCache provides an in-memory cache, delivering sub-millisecond latency for repeatedly requested product descriptions, which directly satisfies the stem's cost-reduction constraint by offloading duplicate foundation model inferences. Unlike persistent stores, its volatile, RAM-based architecture suits transient cached text, avoiding repeated compute charges.

Adding explicit length constraints directly controls the model's output verbosity, which is the stated problem. Unlike vague instructions such as "be brief", specifying "two sentences" gives Amazon Bedrock a concrete target the model can satisfy during generation, tightening the summary without altering the underlying report content or requiring model retraining.

Amazon Bedrock agents can invoke an action group that calls DynamoDB to check current inventory. The agent can fetch in-stock products and either provide them as context to the model or filter the model's recommendations before returning them. This ensures real-time accuracy with minimal latency because DynamoDB provides fast, scalable reads. This is the correct pattern for dynamic inventory validation.

Amazon Lex supplies the conversational interface, handling intent recognition, slot elicitation and dialogue flow, while Amazon Bedrock provides access to foundation models for natural language understanding. Together they satisfy the stem's requirement that two AWS services combine to build a foundation-model-powered chatbot.

PEFT with LoRA freezes the base weights and trains small low-rank adapter matrices, cutting trainable parameters and GPU time substantially. Because original weights remain unchanged, the model's pretrained knowledge is preserved, satisfying both the speed and retention constraints.

The max_tokens parameter specifies the maximum number of tokens to generate in the response. By setting this parameter, the developer can control the length of the summary and ensure it does not exceed the desired token limit. It is a standard parameter supported by Cohere Command on Amazon Bedrock.

Foundation models on Amazon Bedrock can generate product descriptions from keywords using zero-shot prompting. By crafting a clear prompt that specifies the keywords and desired output format, the developer can obtain useful results without any training data or infrastructure management. This is the simplest and most appropriate approach for a prototype with no labeled data.

ListFoundationModels is a Bedrock control-plane action, so the caller's identity must hold bedrock:ListFoundationModels in its attached IAM policy. AccessDeniedException is returned by IAM authorisation evaluation, not by model availability or Region configuration, making a missing or explicitly denied permission the direct cause.

Amazon Bedrock supplies serverless API access to pre-trained foundation models from providers such as Anthropic and Amazon, so no model training or hosting is needed. This directly satisfies the stem's constraint of using a foundation model for sentiment analysis without customisation, unlike services requiring you to build or fine-tune your own model.

RAG grounds responses in the supplied knowledge base by retrieving relevant documents and injecting them into the prompt context, so the model answers from company data rather than its pre-training weights. This directly satisfies the constraint that answers must come only from the internal catalogue.

RAG with an Amazon Bedrock knowledge base retrieves relevant document chunks from the S3 data source at query time and passes them to the model as context. This grounds responses in current content and can return citations, all without retraining or fine-tuning the underlying foundation model.

Stale source content propagates directly into retrieval: Amazon Bedrock knowledge bases sync from the configured data source, so unchanged documents keep returning outdated chunks regardless of model capability. Refreshing or re-syncing the data source satisfies the freshness constraint the stem describes.