Option A provides the most complete, AWS-native defense-in-depth solution for protecting against prompt injection attacks while meeting audit and resiliency requirements. Amazon Bedrock guardrails are designed specifically to enforce safety policies on both user inputs and model outputs, including protections against prompt injection and jailbreak attempts.

Setting content filters to high increases sensitivity to malicious or manipulative inputs. Guardrail profiles allow the same guardrail configuration to be applied consistently across multiple Regions, enabling cross-Region inference and failover without configuration drift. This directly satisfies the requirement for regional resilience.

Amazon CloudWatch Logs captures detailed guardrail intervention events, including when content is blocked, modified, or flagged. Custom metrics derived from these logs enable fine-grained auditing, alerting, and reporting on safety enforcement actions. This provides a more detailed audit trail of safety interventions than API-level logs alone.

Option B adds WAF protection but lacks detailed guardrail intervention logging. Option C introduces additional services and custom logic that increase complexity and may miss model-specific injection patterns. Option D references replication concepts that are not aligned with Bedrock guardrail operational models and relies on word filters, which are insufficient against sophisticated prompt injection techniques.

Therefore, Option A best meets the requirements for layered protection, auditability, and cross-Region resilience using managed Amazon Bedrock safety controls.

Option B best satisfies the requirements while minimizing development effort by combining managed semantic search capabilities with fully managed foundation models. AWS Generative AI guidance describes semantic search as a vector-based retrieval pattern where both documents and user queries are embedded into a shared vector space. Similarity search (such as k-nearest neighbors) then retrieves results based on meaning rather than exact keywords.

Amazon OpenSearch Service natively supports vector indexing and k-NN search at scale. This makes it well suited for large datasets such as 20 million restaurants and 200 million reviews while still achieving sub-second latency for the majority of queries. Because OpenSearch is a distributed, managed service, it automatically scales during peak traffic periods and provides cost-effective performance compared with building and tuning custom vector search pipelines on relational databases.

Using Amazon Bedrock to generate embeddings significantly reduces development complexity. AWS manages the foundation models, eliminates the need for custom model hosting, and ensures consistency by using the same FM for both document embeddings and query embeddings. This aligns directly with AWS-recommended semantic search architectures and removes the need for model lifecycle management.

Hourly updates to restaurant data can be handled efficiently through incremental re-indexing in OpenSearch without disrupting query performance. This approach cleanly separates transactional data storage from search workloads, which is a best practice in AWS architectures.

Option A does not meet the semantic search requirement because keyword-based search cannot reliably interpret complex natural language intent. Option C introduces scalability and performance risks by running large-scale vector similarity searches inside PostgreSQL, which increases operational complexity. Option D adds unnecessary ingestion and abstraction layers intended for retrieval-augmented generation, not high-throughput semantic search.

Therefore, Option B provides the optimal balance of performance, scalability, data freshness, and minimal development effort using AWS Generative AI services.

Option D requires the least development effort because it uses a managed retrieval workflow that bundles the most time-consuming parts of semantic search: embedding generation, vector indexing, and natural language retrieval. With an Amazon Bedrock knowledge base, the application does not need to implement and operate separate services to (1) generate embeddings for hundreds of millions of records, (2) store and manage vectors, (3) build query-time embedding conversion logic, and (4) implement k-NN search orchestration. Instead, the knowledge base is configured to automatically create embeddings during ingestion, and the application queries it using the Amazon Bedrock Retrieve API , which accepts natural language input and performs the vector search as a managed capability.

The performance requirement (95% of queries within 500 ms) is best served by a purpose-built vector search backend rather than running similarity search directly inside a transactional PostgreSQL system at this scale. A knowledge base is designed for retrieval patterns and can be backed by scalable vector stores, which helps meet latency goals under heavy concurrency. The hourly freshness requirement maps naturally to ingestion updates: the pipeline can re-ingest updated restaurant details on a schedule so the knowledge base remains current without building custom re-embedding workflows in application code.

Cost-effective scaling during peak periods is also easier with a managed retrieval layer because scaling the retrieval workload is separated from the operational database. This avoids overprovisioning PostgreSQL for peak semantic-search traffic and reduces the engineering effort to tune performance, sharding, indexing, and retry logic.

Options B and C can work, but they require the team to build and maintain embedding pipelines, query embedding generation, vector index management, and operational scaling strategies. Option A does not provide semantic search because it relies on keyword-based matching rather than embeddings.

Option D is the correct solution because Amazon Q Developer customizations are designed to incorporate organization-approved knowledge and coding guidance without requiring per-project changes. A customization can point Amazon Q Developer to curated internal sources such as approved libraries, coding standards, architectural patterns, and proprietary techniques. This allows the assistant’s suggestions to align with company policies and preferred implementations consistently across teams and repositories.

The key requirement is that the company does not want to make project-level modifications. Options A, B, and C all require adding files or repositories into the project workspace, which directly violates this constraint. They also rely on developer behavior to “use workspace context,” which is harder to enforce and can lead to inconsistent adherence to standards.

With a customization, the organization centrally manages and updates approved resources. This reduces operational overhead because updates to libraries, patterns, or guidelines propagate automatically to developers using the customization, without requiring changes to each project. This is especially valuable for a new team, where consistent enforcement of approved practices is important to reduce compliance risk, security issues, and inconsistent code style.

Additionally, customizations support governance by allowing the company to standardize how Amazon Q Developer responds, ensuring that suggestions reflect approved internal content rather than generic public patterns.

Therefore, Option D best satisfies the requirement for centralized enforcement of approved resources with minimal ongoing management and no project-level modifications.

Option B is the correct solution because it leverages native Amazon Bedrock intelligent prompt routing, which is specifically designed to reduce cost and complexity in multi-model GenAI architectures. Intelligent prompt routing automatically analyzes incoming prompts and selects the most appropriate foundation model based on prompt characteristics and complexity—without requiring custom classification logic or orchestration code.

This approach directly meets the requirement for least implementation effort. The company does not need to deploy additional Lambda functions, maintain routing rules, or manage separate classification stages. Routing decisions are handled by Bedrock, which simplifies architecture and reduces operational risk.

By routing the majority (70%) of simple product inquiries to smaller, lower-cost models, the company minimizes inference cost and latency. More complex return policy inquiries are automatically routed to larger models that provide better reasoning capabilities, preserving response quality and customer satisfaction.

Because routing is handled inline by Bedrock, response latency remains low compared to multi-stage architectures that require an additional classification model call before inference. This is critical for customer service scenarios where responsiveness directly impacts satisfaction.

Option A introduces additional inference steps and custom logic. Option C increases cost by overusing a mid-sized model for all queries. Option D relies on brittle keyword rules and increases operational overhead through endpoint management.

Therefore, Option B delivers the optimal balance of cost efficiency, performance, and simplicity for dynamic model selection in Amazon Bedrock.

Option C is the correct solution because it enforces privacy controls at inference time, not at ingestion time, which is required when different user roles require different visibility into the same underlying data.

Using an S3 Lifecycle configuration ensures that documents older than 3 years are automatically removed, guaranteeing that the knowledge base references only compliant, recent medical reports. Scheduling Lambda-based syncs keeps the knowledge base aligned with the bucket contents without introducing complex per-upload orchestration.

The most important requirement is role-based PII exposure. Amazon Bedrock guardrails support dynamic application at inference time, allowing the system to select a guardrail configuration based on the authenticated user’s Amazon Cognito group. Surgeons can receive full responses, while engineers receive responses with PII masked—without duplicating data or maintaining multiple knowledge bases.

This approach preserves a single source of truth for medical reports while enforcing privacy through response-level controls. It also maintains full auditability of access and redaction behavior.

Option A permanently removes PII and violates surgeon access requirements. Option B redacts data inconsistently and couples privacy logic to ingestion. Option D doubles storage, increases cost, and introduces data drift risk.

Therefore, Option C best meets privacy, compliance, scalability, and operational efficiency requirements.

AWS Step Functions provides native support for human-in-the-loop workflows, making it the best fit for regulatory oversight requirements. The waitForTaskToken integration pattern is explicitly designed to pause a workflow until an external actor—such as a human reviewer—completes a task.

In this architecture, AI-generated recommendations are sent to a human technician for review. The workflow pauses execution using a task token. Once the technician approves or rejects the recommendation, an AWS Lambda function calls SendTaskSuccess or SendTaskFailure, allowing the workflow to continue deterministically.

This approach ensures full auditability, as Step Functions records every state transition, timestamp, and execution path. Storing review outcomes in Amazon DynamoDB provides durable, queryable audit records required for regulatory compliance.

Option A requires custom orchestration and lacks native workflow state management. Option C incorrectly uses AWS Glue, which is not designed for approval workflows. Option D uses caching instead of durable audit storage and introduces unnecessary complexity.

Therefore, Option B is the AWS-recommended, lowest-risk, and most auditable solution for mandatory human review of AI outputs.

Amazon SageMaker Serverless Inference is specifically designed for applications that experience intermittent or bursty traffic. It automatically scales compute capacity based on the number of requests and scales down to zero when there is no traffic, satisfying the requirement to optimize resource usage during low demand. To meet the 500 ms latency requirement during peak periods and avoid " cold start " delays, provisioned concurrency keeps a specified number of execution environments warm and ready to respond immediately. This provides a balance between the cost-effectiveness of serverless and the performance predictability of provisioned instances. Multi-model endpoints (Option A) can introduce " noisy neighbor " issues and latency spikes, while asynchronous inference (Option D) is intended for long-running workloads and cannot meet sub-500 ms requirements.

Option A is the correct solution because it relies on native Amazon Bedrock capabilities to deliver transparency, auditability, scalability, and low latency with minimal operational overhead. Amazon Bedrock Knowledge Bases provide a fully managed Retrieval Augmented Generation (RAG) implementation that automatically handles document ingestion, embedding, retrieval, and source attribution, enabling the application to cite authoritative content without building custom pipelines.

Enabling tracing for Amazon Bedrock Agents provides end-to-end visibility into agent reasoning steps, tool usage, and model interactions. This satisfies the requirement for comprehensive audit trails and supports regulatory review in financial services environments. Structured prompts further ensure that responses explicitly present reasoning and supporting evidence in a controlled, auditable format.

Using Amazon API Gateway and AWS Lambda allows the application to scale automatically to thousands of concurrent users without capacity planning. These services are designed for bursty workloads and can easily support the stated requirement of up to 10,000 concurrent users. Amazon CloudFront reduces latency by caching and accelerating content delivery, helping the application meet the strict 2-second response-time requirement.

Option B introduces a custom RAG pipeline with OpenSearch, increasing operational complexity and maintenance effort. Option C lacks native RAG integration and does not provide transparent reasoning or citation management. Option D focuses on offline compliance reporting rather than real-time transparency and low-latency responses.

Therefore, Option A best meets all requirements while minimizing infrastructure and operational overhead.

Option C best satisfies the requirement to change routing decisions without redeploying code while supporting complex, frequently changing business logic at scale. AWS AppConfig is designed for centrally managing dynamic configuration (feature flags, rules, thresholds, and policy parameters) and deploying changes safely. It supports controlled deployments, validation, and rapid propagation of updated configuration values, which aligns with “real-time cost metrics that change hourly” and the need for “immediate propagation across thousands of concurrent requests.”

In this design, the Lambda function becomes the policy decision point. For each request, it evaluates user attributes (tier, transaction value), context (regulatory zone, Region), and live cost/performance thresholds stored in AppConfig to determine which Amazon Bedrock FM to invoke. Because the routing rules and FM identifiers are delivered as configuration, the company can switch models, adjust A/B testing weights, or update compliance routing rules by deploying new AppConfig configuration versions rather than pushing new application code. This reduces operational risk and accelerates iteration.

Exposing a single API Gateway endpoint also minimizes client complexity and keeps routing logic server-side, which is important when rules change frequently. Lambda can cache configuration between invocations (within the execution environment) to reduce repeated fetch overhead while still picking up changes quickly, enabling both low latency and rapid rule rollout under high concurrency.

Option A relies on Lambda environment variables, which are not intended for frequent real-time updates and typically require function configuration updates that are slower and operationally brittle. Option B uses mapping templates and stage variables, which are limited for complex rule evaluation and safe rollout patterns. Option D misuses authorizers for business routing, adds extra latency and complexity, and complicates observability and error handling by splitting decisioning from execution.