Cisco 300-640 Practice Test 4
Question 30
The acceptance board must decide whether an AI-infrastructure response to interactive inference slos is fit for controlled operation. The fictional operating facts are: A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. Signoff cannot rely on a good-looking utilization graph, a change ticket alone or an unmeasured promise. The response must include an appropriate governing engineering principle, a scoped intervention and a separate measurable acceptance criterion. Choose EXACTLY THREE statements that meet those roles.
- A staged test proves authorized transfer, version parity and the documented loss-of-connectivity behavior.
- Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization.
- Establish least-privilege hybrid links, versioned artifact replication and fail-closed handling for stale data or disconnected sites.
- Choose an AI use case only after defining the decision, reliable labels or feedback, error costs, human controls and a non-AI baseline.
- The approved peak test meets p99 response latency and error-rate budgets without exhausting accelerator memory.
- Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
Explanation
Correct Answer
B:A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization. Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
E:A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. The approved peak test meets p99 response latency and error-rate budgets without exhausting accelerator memory. Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
F:A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective. Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization.
Incorrect Answers
A:A staged test proves authorized transfer, version parity and the documented loss-of-connectivity behavior. It would not independently demonstrate that the control works: A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. The approved peak test meets p99 response latency and error-rate budgets without exhausting accelerator memory. Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
C:Establish least-privilege hybrid links, versioned artifact replication and fail-closed handling for stale data or disconnected sites. It does not implement the governing rule for this approval: A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective. The approved peak test meets p99 response latency and error-rate budgets without exhausting accelerator memory.
D:Choose an AI use case only after defining the decision, reliable labels or feedback, error costs, human controls and a non-AI baseline. It is not the rule that supports the proposed change and acceptance test: A multilingual chatbot rolls out larger batches and reports higher GPU throughput while its busiest customer conversations regularly time out. Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization. Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
Question 31
A data-center AI architecture committee considers controlled model lifecycle promotion in its AI Fundamentals and Applications review. At the affected site: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. The committee is deliberately not selecting a vendor product or applying a live change yet. It needs the foundational reasoning that makes the next design decision defensible. Which ONE engineering principle directly governs this situation? Choose ONE answer.
- Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization.
- Workload placement must satisfy binding residency, data-handling and latency constraints before cloud-versus-on-premises cost comparisons.
- A hybrid workload needs explicit identity, encrypted transport, consistent model and data versions, and ownership of cross-site failure modes.
- Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Explanation
Correct Answer
D:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
Incorrect Answers
A:Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization. It does not govern the condition described: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
B:Workload placement must satisfy binding residency, data-handling and latency constraints before cloud-versus-on-premises cost comparisons. It does not govern the condition described: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
C:A hybrid workload needs explicit identity, encrypted transport, consistent model and data versions, and ownership of cross-site failure modes. It does not govern the condition described: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Question 32
An infrastructure diagnostician is investigating controlled model lifecycle promotion for the AI Fundamentals and Applications team. The reported incident reads: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Competing teams attribute the symptoms to different layers of the AI stack. Before accepting a hypothesis or touching production settings, the investigator must choose a discriminating set of measurements. Which ONE evidence package is MOST relevant? Choose ONE answer.
- Map data categories, access jurisdiction, egress paths, encryption-key custody, hardware availability and required response latency.
- Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
- Record on-device model footprint, local input rates, offline duration, update signatures and safe fallback triggers.
- Inspect pod resource requests, device-plugin visibility, node labels, quota, volume readiness and scheduling events.
Explanation
Correct Answer
B:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Incorrect Answers
A:Map data categories, access jurisdiction, egress paths, encryption-key custody, hardware availability and required response latency. That check would not establish the decisive facts for this condition: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
C:Record on-device model footprint, local input rates, offline duration, update signatures and safe fallback triggers. That check would not establish the decisive facts for this condition: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
D:Inspect pod resource requests, device-plugin visibility, node labels, quota, volume readiness and scheduling events. That check would not establish the decisive facts for this condition: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
Question 33
During a bounded AI-platform change-control review concerning controlled model lifecycle promotion, the site owner reports: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. The proposed work must have a narrow implementation scope, first be validated against the failure evidence and avoid unrelated system-wide tuning. Which ONE intervention should the engineers pilot before any wider rollout? Choose ONE answer.
- Define hardware-aware scheduling constraints, namespaces and resource quotas with documented preemption and restart behavior.
- Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
- Correct document permissions and retrieval filters, refresh the stale index and evaluate grounded responses on representative queries.
- Deploy a compact validated inference artifact with signed updates, bounded local buffering and a defined offline fallback.
Explanation
Correct Answer
B:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
Incorrect Answers
A:Define hardware-aware scheduling constraints, namespaces and resource quotas with documented preemption and restart behavior. That change would leave the reported problem unresolved: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
C:Correct document permissions and retrieval filters, refresh the stale index and evaluate grounded responses on representative queries. That change would leave the reported problem unresolved: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
D:Deploy a compact validated inference artifact with signed updates, bounded local buffering and a defined offline fallback. That change would leave the reported problem unresolved: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
Question 34
Two independent technical reviewers examine a AI Fundamentals and Applications problem involving controlled model lifecycle promotion. Their shared incident log states: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. One reviewer requires measurements capable of separating the observed problem from neighboring failure modes; the other requires a bounded action that follows those measurements. Choose EXACTLY TWO statements that together supply those two parts.
- Measure all-reduce duration, PCIe and NVLink topology, GPU locality, host-to-network contention and the supported platform matrix.
- Map data categories, access jurisdiction, egress paths, encryption-key custody, hardware availability and required response latency.
- Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
- Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
- Match the training and checkpoint workflow to supported file or block protocols and benchmark access under representative concurrency.
Explanation
Correct Answer
C:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
D:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
Incorrect Answers
A:Measure all-reduce duration, PCIe and NVLink topology, GPU locality, host-to-network contention and the supported platform matrix. It would not provide the evidence needed for the paired diagnosis and follow-up: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
B:Map data categories, access jurisdiction, egress paths, encryption-key custody, hardware availability and required response latency. It would not provide the evidence needed for the paired diagnosis and follow-up: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
E:Match the training and checkpoint workflow to supported file or block protocols and benchmark access under representative concurrency. It is not the scoped response supported by the accompanying evidence: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
Question 35
The acceptance board must decide whether an AI-infrastructure response to controlled model lifecycle promotion is fit for controlled operation. The fictional operating facts are: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Signoff cannot rely on a good-looking utilization graph, a change ticket alone or an unmeasured promise. The response must include an appropriate governing engineering principle, a scoped intervention and a separate measurable acceptance criterion. Choose EXACTLY THREE statements that meet those roles.
- Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
- Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
- Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
- Data-flow review shows no unauthorized cross-boundary movement and the local workload meets its latency target.
- Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
- Choose an AI use case only after defining the decision, reliable labels or feedback, error costs, human controls and a non-AI baseline.
Explanation
Correct Answer
B:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
C:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
E:A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
Incorrect Answers
A:Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective. It does not implement the governing rule for this approval: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
D:Data-flow review shows no unauthorized cross-boundary movement and the local workload meets its latency target. It would not independently demonstrate that the control works: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
F:Choose an AI use case only after defining the decision, reliable labels or feedback, error costs, human controls and a non-AI baseline. It is not the rule that supports the proposed change and acceptance test: A demand-prediction service suddenly increases overstock after a model refresh but its team cannot link the serving image to an evaluation run. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
Question 36
A data-center AI architecture committee considers controlled model lifecycle promotion in its AI Fundamentals and Applications review. At the affected site: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. The committee is deliberately not selecting a vendor product or applying a live change yet. It needs the foundational reasoning that makes the next design decision defensible. Which ONE engineering principle directly governs this situation? Choose ONE answer.
- Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization.
- Model training modifies learned parameters using a controlled training pipeline; serving a model for inference should not silently retrain it on each user request.
- A retrieval-augmented assistant should fetch authorized current passages and ground its responses in those passages rather than rely on model memory alone.
- Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Explanation
Correct Answer
D:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
Incorrect Answers
A:Latency-sensitive inference capacity must be sized for peak request arrival, model memory, batching and tail response time, not merely average accelerator utilization. It does not govern the condition described: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
B:Model training modifies learned parameters using a controlled training pipeline; serving a model for inference should not silently retrain it on each user request. It does not govern the condition described: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
C:A retrieval-augmented assistant should fetch authorized current passages and ground its responses in those passages rather than rely on model memory alone. It does not govern the condition described: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Question 37
An infrastructure diagnostician is investigating controlled model lifecycle promotion for the AI Fundamentals and Applications team. The reported incident reads: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Competing teams attribute the symptoms to different layers of the AI stack. Before accepting a hypothesis or touching production settings, the investigator must choose a discriminating set of measurements. Which ONE evidence package is MOST relevant? Choose ONE answer.
- Review official supported components, approved release matrix, topology, GPU-network-storage bottlenecks and site facilities.
- Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
- Inspect pod resource requests, device-plugin visibility, node labels, quota, volume readiness and scheduling events.
- Record on-device model footprint, local input rates, offline duration, update signatures and safe fallback triggers.
Explanation
Correct Answer
B:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Incorrect Answers
A:Review official supported components, approved release matrix, topology, GPU-network-storage bottlenecks and site facilities. That check would not establish the decisive facts for this condition: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
C:Inspect pod resource requests, device-plugin visibility, node labels, quota, volume readiness and scheduling events. That check would not establish the decisive facts for this condition: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
D:Record on-device model footprint, local input rates, offline duration, update signatures and safe fallback triggers. That check would not establish the decisive facts for this condition: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
Question 38
During a bounded AI-platform change-control review concerning controlled model lifecycle promotion, the site owner reports: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. The proposed work must have a narrow implementation scope, first be validated against the failure evidence and avoid unrelated system-wide tuning. Which ONE intervention should the engineers pilot before any wider rollout? Choose ONE answer.
- Pilot the narrowly scoped inference workflow with a named owner and benchmark it against a simple deterministic baseline.
- Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
- Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective.
- Compare the planned cluster with the current validated Cisco reference architecture and test deviations in a controlled pilot.
Explanation
Correct Answer
B:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
Incorrect Answers
A:Pilot the narrowly scoped inference workflow with a named owner and benchmark it against a simple deterministic baseline. That change would leave the reported problem unresolved: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
C:Separate interactive serving capacity from bulk training work and tune batching or autoscaling against an explicit tail-latency objective. That change would leave the reported problem unresolved: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
D:Compare the planned cluster with the current validated Cisco reference architecture and test deviations in a controlled pilot. That change would leave the reported problem unresolved: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
Question 39
Two independent technical reviewers examine a AI Fundamentals and Applications problem involving controlled model lifecycle promotion. Their shared incident log states: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. One reviewer requires measurements capable of separating the observed problem from neighboring failure modes; the other requires a bounded action that follows those measurements. Choose EXACTLY TWO statements that together supply those two parts.
- Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals.
- Compare the user's query with retrieved document IDs, access scopes, citation coverage and the freshness of indexed versions.
- Place tightly coupled workers on topology-compatible GPU groups and validate communication scaling before expanding nodes.
- Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
- Map data categories, access jurisdiction, egress paths, encryption-key custody, hardware availability and required response latency.
Explanation
Correct Answer
A:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
D:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
Incorrect Answers
B:Compare the user's query with retrieved document IDs, access scopes, citation coverage and the freshness of indexed versions. It would not provide the evidence needed for the paired diagnosis and follow-up: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
C:Place tightly coupled workers on topology-compatible GPU groups and validate communication scaling before expanding nodes. It is not the scoped response supported by the accompanying evidence: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
E:Map data categories, access jurisdiction, egress paths, encryption-key custody, hardware availability and required response latency. It would not provide the evidence needed for the paired diagnosis and follow-up: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Trace dataset hash, training run, evaluation results, model registry tag, deployed image and production drift signals. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
Question 40
The acceptance board must decide whether an AI-infrastructure response to controlled model lifecycle promotion is fit for controlled operation. The fictional operating facts are: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Signoff cannot rely on a good-looking utilization graph, a change ticket alone or an unmeasured promise. The response must include an appropriate governing engineering principle, a scoped intervention and a separate measurable acceptance criterion. Choose EXACTLY THREE statements that meet those roles.
- Choose an AI use case only after defining the decision, reliable labels or feedback, error costs, human controls and a non-AI baseline.
- Data loaders sustain the target sample rate while checkpoints restore correctly under a simulated host loss.
- Match the training and checkpoint workflow to supported file or block protocols and benchmark access under representative concurrency.
- Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.
- Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
- Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
Explanation
Correct Answer
D:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
E:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers.
F:A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
Incorrect Answers
A:Choose an AI use case only after defining the decision, reliable labels or feedback, error costs, human controls and a non-AI baseline. It is not the rule that supports the proposed change and acceptance test: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Reliable AI delivery separates data validation, training, evaluation, artifact registry, deployment, telemetry and rollback with reproducible version identifiers. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
B:Data loaders sustain the target sample rate while checkpoints restore correctly under a simulated host loss. It would not independently demonstrate that the control works: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version.
C:Match the training and checkpoint workflow to supported file or block protocols and benchmark access under representative concurrency. It does not implement the governing rule for this approval: A document-classification product promotes an apparently improved model without testing a new language distribution and lacks a rollback checkpoint. Gate model promotion on evaluation and policy checks, retain lineage and provide a tested rollback to the preceding known-good version. Operators can identify the exact dataset-model-image combination and restore the approved baseline in a drill.