cert
cert-1
cert-2

Pass Microsoft AI-300 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
AI-300 Exam - Verified By Experts
AI-300 Premium File

AI-300 Premium File

$69.99
$76.99
  • Premium File 213 Questions & Answers. Last Update: Nov 12, 2026

Whats Included:

  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates
 
$76.99
$69.99
accept 45 downloads in the last 7 days
block-screenshots
Microsoft AI-300 Practice Test 3

Question 20

As an AI operations team checks production agent evaluation evidence, the following condition occurs: A user-facing agent becomes slow after an LLM deployment change, but the team only records total response time and cannot tell retrieval delay from model generation delay. The workflow must produce measurable quality, safety and performance evidence. Which diagnostic evidence most directly distinguishes the described AI operations failure?

  1. Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.
  2. Inspect monitoring sample filters, event statuses, language distribution, evaluator eligibility and collected trace metadata.
  3. Review usage logs for chat and embedding deployments, prompt/output token categories, query frequency, indexing cycles and resource billing tags.
  4. Examine end-to-end request traces, retrieval timing, model time to first token, generation duration, request concurrency and percentiles.

Explanation

Correct Answer

D:

Examine end-to-end request traces, retrieval timing, model time to first token, generation duration, request concurrency and percentiles. This is the most relevant evidence for generation latency spike blamed on retrieval without breakdown. Latency budgets should separate agent orchestration, retrieval and generation; total duration alone cannot locate the bottleneck. The matching controlled remediation is: Instrument stage-level latency and compare generation and retrieval spans before applying a targeted capacity or retrieval change.

Incorrect Answers

A:

Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations. This would investigate multi-turn agent evaluation uses isolated last message only rather than the documented generation latency spike blamed on retrieval without breakdown issue. The relevant investigation is: Examine end-to-end request traces, retrieval timing, model time to first token, generation duration, request concurrency and percentiles.

B:

Inspect monitoring sample filters, event statuses, language distribution, evaluator eligibility and collected trace metadata. This would investigate continuous monitoring has no representative sampling coverage rather than the documented generation latency spike blamed on retrieval without breakdown issue. The relevant investigation is: Examine end-to-end request traces, retrieval timing, model time to first token, generation duration, request concurrency and percentiles.

C:

Review usage logs for chat and embedding deployments, prompt/output token categories, query frequency, indexing cycles and resource billing tags. This would investigate token cost report omits embedding requests rather than the documented generation latency spike blamed on retrieval without breakdown issue. The relevant investigation is: Examine end-to-end request traces, retrieval timing, model time to first token, generation duration, request concurrency and percentiles.

Question 21

During a retrieval or model-customization performance review, the following condition occurs: A compliance RAG application retrieves semantically similar policies but fails queries containing an exact statute or clause identifier because the index uses vector-only retrieval. The change must keep citations, privacy and previously approved behavior intact. Select TWO actions: one decisive diagnostic and its corresponding targeted remediation.

  1. Use a governed retrieval-augmented grounding path for volatile factual content and reserve fine-tuning for stable behavior or output-format adaptation.
  2. Use an appropriate hybrid keyword-plus-vector search strategy with evaluated fusion and optional semantic reranking.
  3. Filter or downrank superseded records before hybrid retrieval and use governed version metadata for source selection.
  4. Inspect representative code-based queries, BM25/keyword results, vector matches, analyzer configuration and fusion ranking behavior.
  5. Examine question-set provenance, tuning history, held-out set separation, document version and evaluator configuration.

Explanation

Correct Answers

B:

Use an appropriate hybrid keyword-plus-vector search strategy with evaluated fusion and optional semantic reranking. This targets pure vector search misses exact regulatory clause codes. Semantic vectors are strong at meaning but exact identifiers often benefit from lexical search; hybrid retrieval can serve both needs. Verify the failure through: Inspect representative code-based queries, BM25/keyword results, vector matches, analyzer configuration and fusion ranking behavior. Validate success with: Both natural-language and exact-identifier questions retrieve the correct authoritative clause without increasing unrelated matches.

D:

Inspect representative code-based queries, BM25/keyword results, vector matches, analyzer configuration and fusion ranking behavior. This is the most relevant evidence for pure vector search misses exact regulatory clause codes. Semantic vectors are strong at meaning but exact identifiers often benefit from lexical search; hybrid retrieval can serve both needs. The matching controlled remediation is: Use an appropriate hybrid keyword-plus-vector search strategy with evaluated fusion and optional semantic reranking.

Incorrect Answers

A:

Use a governed retrieval-augmented grounding path for volatile factual content and reserve fine-tuning for stable behavior or output-format adaptation. This is associated with fine-tuning task selected when retrieval would better handle changing facts, not the essential evidence-and-fix pair for pure vector search misses exact regulatory clause codes. First examine: Inspect representative code-based queries, BM25/keyword results, vector matches, analyzer configuration and fusion ranking behavior. Then correct: Use an appropriate hybrid keyword-plus-vector search strategy with evaluated fusion and optional semantic reranking.

C:

Filter or downrank superseded records before hybrid retrieval and use governed version metadata for source selection. This is associated with hybrid retrieval double-counts outdated documents, not the essential evidence-and-fix pair for pure vector search misses exact regulatory clause codes. First examine: Inspect representative code-based queries, BM25/keyword results, vector matches, analyzer configuration and fusion ranking behavior. Then correct: Use an appropriate hybrid keyword-plus-vector search strategy with evaluated fusion and optional semantic reranking.

E:

Examine question-set provenance, tuning history, held-out set separation, document version and evaluator configuration. This is associated with offline rag evaluator leaks training questions into test set, not the essential evidence-and-fix pair for pure vector search misses exact regulatory clause codes. First examine: Inspect representative code-based queries, BM25/keyword results, vector matches, analyzer configuration and fusion ranking behavior. Then correct: Use an appropriate hybrid keyword-plus-vector search strategy with evaluated fusion and optional semantic reranking.

Question 22

During a retrieval or model-customization performance review, the following condition occurs: An operations chatbot answers 'I cannot find it' to precise internal procedure questions because its vector search similarity threshold was raised above scores returned by the relevant documents. The change must keep citations, privacy and previously approved behavior intact. Select TWO actions: one decisive diagnostic and its corresponding targeted remediation.

  1. Calibrate retrieval threshold and ranking settings using labeled queries to admit relevant context without overwhelming the model with noise.
  2. Check document effective dates, index filters, scoring profiles, vector/keyword rankings and available canonical document metadata.
  3. Re-embed and reindex the document corpus with a compatible versioned embedding model, or keep query embeddings pinned to the existing index model.
  4. Refresh a governed evaluation set with recent terminology and trigger controlled retraining only after measuring sustained quality regression.
  5. Inspect top-k search scores, threshold, query normalization, relevant passages and test question recall.

Explanation

Correct Answers

A:

Calibrate retrieval threshold and ranking settings using labeled queries to admit relevant context without overwhelming the model with noise. This targets rag similarity threshold filters all relevant evidence. Vector similarity thresholds control recall and precision; a threshold chosen without validation can suppress even the best available evidence. Verify the failure through: Inspect top-k search scores, threshold, query normalization, relevant passages and test question recall. Validate success with: Previously missed procedure passages are retrieved and grounded answer quality improves without a large irrelevant context increase.

E:

Inspect top-k search scores, threshold, query normalization, relevant passages and test question recall. This is the most relevant evidence for rag similarity threshold filters all relevant evidence. Vector similarity thresholds control recall and precision; a threshold chosen without validation can suppress even the best available evidence. The matching controlled remediation is: Calibrate retrieval threshold and ranking settings using labeled queries to admit relevant context without overwhelming the model with noise.

Incorrect Answers

B:

Check document effective dates, index filters, scoring profiles, vector/keyword rankings and available canonical document metadata. This is associated with hybrid retrieval double-counts outdated documents, not the essential evidence-and-fix pair for rag similarity threshold filters all relevant evidence. First examine: Inspect top-k search scores, threshold, query normalization, relevant passages and test question recall. Then correct: Calibrate retrieval threshold and ranking settings using labeled queries to admit relevant context without overwhelming the model with noise.

C:

Re-embed and reindex the document corpus with a compatible versioned embedding model, or keep query embeddings pinned to the existing index model. This is associated with embedding model replaced without reindexing stored vectors, not the essential evidence-and-fix pair for rag similarity threshold filters all relevant evidence. First examine: Inspect top-k search scores, threshold, query normalization, relevant passages and test question recall. Then correct: Calibrate retrieval threshold and ranking settings using labeled queries to admit relevant context without overwhelming the model with noise.

D:

Refresh a governed evaluation set with recent terminology and trigger controlled retraining only after measuring sustained quality regression. This is associated with fine-tune evaluation lacks drift checks for new customer terminology, not the essential evidence-and-fix pair for rag similarity threshold filters all relevant evidence. First examine: Inspect top-k search scores, threshold, query normalization, relevant passages and test question recall. Then correct: Calibrate retrieval threshold and ranking settings using labeled queries to admit relevant context without overwhelming the model with noise.

Question 23

During a production Azure Machine Learning model lifecycle incident, the following condition occurs: A tuning experiment requests dozens of concurrent GPU trials with no bound, leaving the Azure ML workspace queued for hours under regional quota constraints. The operator must safeguard training reproducibility and serving continuity. Which TWO complementary choices identify the issue and correct it? Select TWO.

  1. Configure a suitable seasonal baseline or context-aware drift monitor while retaining meaningful anomaly thresholds.
  2. Inspect sweep max_concurrent_trials, compute cluster maximum nodes, GPU family quota and job scheduling telemetry.
  3. Capture a governed immutable dataset or versioned data asset and reference that exact version in notebook and scheduled jobs.
  4. Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process.
  5. Set justified trial concurrency and sampling/early-termination limits so the sweep fits available GPU capacity.

Explanation

Correct Answers

B:

Inspect sweep max_concurrent_trials, compute cluster maximum nodes, GPU family quota and job scheduling telemetry. This is the most relevant evidence for sweep concurrency overwhelms scarce gpu quota. Hyperparameter tuning parallelism and cloud compute quota must be planned together; more requested trials do not create physical capacity. The matching controlled remediation is: Set justified trial concurrency and sampling/early-termination limits so the sweep fits available GPU capacity.

E:

Set justified trial concurrency and sampling/early-termination limits so the sweep fits available GPU capacity. This targets sweep concurrency overwhelms scarce gpu quota. Hyperparameter tuning parallelism and cloud compute quota must be planned together; more requested trials do not create physical capacity. Verify the failure through: Inspect sweep max_concurrent_trials, compute cluster maximum nodes, GPU family quota and job scheduling telemetry. Validate success with: Trials execute predictably without persistent quota-induced queueing, and tuning still explores the intended search space.

Incorrect Answers

A:

Configure a suitable seasonal baseline or context-aware drift monitor while retaining meaningful anomaly thresholds. This is associated with data drift monitor uses wrong reference season, not the essential evidence-and-fix pair for sweep concurrency overwhelms scarce gpu quota. First examine: Inspect sweep max_concurrent_trials, compute cluster maximum nodes, GPU family quota and job scheduling telemetry. Then correct: Set justified trial concurrency and sampling/early-termination limits so the sweep fits available GPU capacity.

C:

Capture a governed immutable dataset or versioned data asset and reference that exact version in notebook and scheduled jobs. This is associated with exploration notebook reads production data without reproducible snapshot, not the essential evidence-and-fix pair for sweep concurrency overwhelms scarce gpu quota. First examine: Inspect sweep max_concurrent_trials, compute cluster maximum nodes, GPU family quota and job scheduling telemetry. Then correct: Set justified trial concurrency and sampling/early-termination limits so the sweep fits available GPU capacity.

D:

Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process. This is associated with mlflow tracking uri sends notebook runs to wrong workspace, not the essential evidence-and-fix pair for sweep concurrency overwhelms scarce gpu quota. First examine: Inspect sweep max_concurrent_trials, compute cluster maximum nodes, GPU family quota and job scheduling telemetry. Then correct: Set justified trial concurrency and sampling/early-termination limits so the sweep fits available GPU capacity.

Question 24

During a Foundry generative AI quality and observability incident, the following condition occurs: A retrieval assistant generates fluent answers citing unrelated documents, yet the release dashboard declares success based solely on readability. The fix must preserve evaluation validity and avoid unnecessary user-data collection. What measured outcome validates the solution while preserving the stated constraints?

  1. An induced tool timeout is attributable to the exact trace span and remediation can target the failing service.
  2. Unsupported claims fail the release gate even when the answers are grammatically smooth.
  3. The evaluation reveals and measures failures that depend on follow-up context and the selected agent version improves those cases.
  4. Historical baselines can be reproduced with the same evaluator version and new rubric results are labeled separately.

Explanation

Correct Answer

B:

Unsupported claims fail the release gate even when the answers are grammatically smooth. Readability and factual support are distinct dimensions of GenAI output quality; polished prose is not evidence that retrieved sources justify a claim. a direct corrective action is to add context-groundedness and relevance evaluation on representative answers alongside coherence and fluency.

Incorrect Answers

A:

An induced tool timeout is attributable to the exact trace span and remediation can target the failing service. That addresses a different configuration issue rather than the main requirement. Verify instead: Unsupported claims fail the release gate even when the answers are grammatically smooth.

C:

The evaluation reveals and measures failures that depend on follow-up context and the selected agent version improves those cases. That addresses a different configuration issue rather than the main requirement. Verify instead: Unsupported claims fail the release gate even when the answers are grammatically smooth.

D:

Historical baselines can be reproduced with the same evaluator version and new rubric results are labeled separately. That addresses a different configuration issue rather than the main requirement. Verify instead: Unsupported claims fail the release gate even when the answers are grammatically smooth.

Question 25

During a Microsoft Foundry generative AI deployment review, the following condition occurs: A legal-document summarizer routinely receives 110,000-token inputs, but the selected model deployment has an insufficient supported context window and rejects or truncates requests. The deployment must retain governed identities, endpoint contracts and auditable release history. What approved correction resolves the failure without modifying unrelated healthy services?

  1. Preflight capacity and schedule an approved upgrade with quota adjustment or temporary side-by-side resources that fit the region.
  2. Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements.
  3. Deploy an available supported online/serverless or managed model endpoint sized for synchronous conversational requests.
  4. Separate trusted instructions from quoted retrieved context and enforce restricted tool permissions and injection-resilient prompt handling.

Explanation

Correct Answer

B:

Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements. This targets foundation model selected without required context window. Model suitability depends on context limits and task capability, not just a broad 'generative model' label. Verify the failure through: Compare real tokenized input sizes, selected model version limits, deployment configuration and representative failure logs. Validate success with: A production-sized document completes evaluation without unintended context loss or unsupported token-limit errors.

Incorrect Answers

A:

Preflight capacity and schedule an approved upgrade with quota adjustment or temporary side-by-side resources that fit the region. This could address provisioned model update exceeds regional capacity ceiling, but it does not solve the specified foundation model selected without required context window failure. The matching remediation is: Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements.

C:

Deploy an available supported online/serverless or managed model endpoint sized for synchronous conversational requests. This could address interactive inference deployed on unsuitable batch-style compute, but it does not solve the specified foundation model selected without required context window failure. The matching remediation is: Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements.

D:

Separate trusted instructions from quoted retrieved context and enforce restricted tool permissions and injection-resilient prompt handling. This could address prompt includes untrusted retrieved instructions as system policy, but it does not solve the specified foundation model selected without required context window failure. The matching remediation is: Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements.

Question 26

As an AI operations team checks production agent evaluation evidence, the following condition occurs: A GenAIOps release pipeline writes evaluation scores to a report but deploys an agent even when groundedness falls below the required threshold. The workflow must produce measurable quality, safety and performance evidence. Which investigation best verifies the cause before changing production configuration?

  1. Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs.
  2. Inspect evaluation dataset field names, input-output column mapping, evaluator requirements, row samples and failed scoring logs.
  3. Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.
  4. Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records.

Explanation

Correct Answer

D:

Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. This is the most relevant evidence for evaluation workflow fails to block degraded agent deployment. Automated evaluation provides protection only when results are connected to enforceable deployment decisions. The matching controlled remediation is: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.

Incorrect Answers

A:

Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs. This would investigate agent traces omit downstream tool failure reason rather than the documented evaluation workflow fails to block degraded agent deployment issue. The relevant investigation is: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records.

B:

Inspect evaluation dataset field names, input-output column mapping, evaluator requirements, row samples and failed scoring logs. This would investigate genai evaluation dataset lacks expected answer column mapping rather than the documented evaluation workflow fails to block degraded agent deployment issue. The relevant investigation is: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records.

C:

Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations. This would investigate multi-turn agent evaluation uses isolated last message only rather than the documented evaluation workflow fails to block degraded agent deployment issue. The relevant investigation is: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records.

Question 27

As an AI operations team checks production agent evaluation evidence, the following condition occurs: A Foundry agent returns an unhelpful answer after a search tool timeout but logs only the final generated text, hiding the actual failing service and request correlation. The workflow must produce measurable quality, safety and performance evidence. Which TWO complementary choices identify the issue and correct it? Select TWO.

  1. Inspect prompt token composition, repeated static sections, retrieval result counts, cache opportunities and groundedness metrics.
  2. Enable scoped distributed tracing for agent and tool steps with error tags and privacy-safe identifiers.
  3. Extend safety testing to representative supported languages and calibrate policy thresholds without weakening necessary allowed content.
  4. Mask sensitive tokens and personal data, apply least-privilege log access and retain only necessary structured debugging evidence.
  5. Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs.

Explanation

Correct Answers

B:

Enable scoped distributed tracing for agent and tool steps with error tags and privacy-safe identifiers. This targets agent traces omit downstream tool failure reason. Multi-step AI agent troubleshooting needs per-tool traces, not just the final answer and an overall request status. Verify the failure through: Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs. Validate success with: An induced tool timeout is attributable to the exact trace span and remediation can target the failing service.

E:

Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs. This is the most relevant evidence for agent traces omit downstream tool failure reason. Multi-step AI agent troubleshooting needs per-tool traces, not just the final answer and an overall request status. The matching controlled remediation is: Enable scoped distributed tracing for agent and tool steps with error tags and privacy-safe identifiers.

Incorrect Answers

A:

Inspect prompt token composition, repeated static sections, retrieval result counts, cache opportunities and groundedness metrics. This is associated with prompt truncation strategy increases generated-token spend, not the essential evidence-and-fix pair for agent traces omit downstream tool failure reason. First examine: Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs. Then correct: Enable scoped distributed tracing for agent and tool steps with error tags and privacy-safe identifiers.

C:

Extend safety testing to representative supported languages and calibrate policy thresholds without weakening necessary allowed content. This is associated with harmful content probe set excludes multilingual input, not the essential evidence-and-fix pair for agent traces omit downstream tool failure reason. First examine: Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs. Then correct: Enable scoped distributed tracing for agent and tool steps with error tags and privacy-safe identifiers.

D:

Mask sensitive tokens and personal data, apply least-privilege log access and retain only necessary structured debugging evidence. This is associated with debug traces leak secrets and retrieved personal records, not the essential evidence-and-fix pair for agent traces omit downstream tool failure reason. First examine: Inspect agent instrumentation, OpenTelemetry trace propagation, tool span durations, error attributes and correlated resource logs. Then correct: Enable scoped distributed tracing for agent and tool steps with error tags and privacy-safe identifiers.

Question 28

During a production Azure Machine Learning model lifecycle incident, the following condition occurs: A multi-node PyTorch training job starts on Azure ML GPUs but stalls during collective initialization because the managed network blocks approved inter-node rendezvous communication. The operator must safeguard training reproducibility and serving continuity. Which production-safe remediation should the engineer apply to this AI workflow?

  1. Implement feasible deterministic seeding and record input data split and environment versions for each run with acknowledged nondeterministic GPU constraints.
  2. Write the model into the declared Azure ML job output path and track the versioned artifact used in registration.
  3. Allow the narrowly scoped supported compute-to-compute communication paths and correct the distributed training rendezvous configuration.
  4. Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.

Explanation

Correct Answer

C:

Allow the narrowly scoped supported compute-to-compute communication paths and correct the distributed training rendezvous configuration. This targets distributed job workers cannot communicate over required network path. Distributed training requires connectivity and consistent rank/world-size setup among workers; GPU availability alone does not establish a working cluster. Verify the failure through: Inspect distributed job settings, node count, rendezvous endpoint configuration, compute network security rules and worker log timeouts. Validate success with: All workers join one process group and finish a representative multi-node training step without broad public inbound exposure.

Incorrect Answers

A:

Implement feasible deterministic seeding and record input data split and environment versions for each run with acknowledged nondeterministic GPU constraints. This could address training script ignores reproducible seed and input contract, but it does not solve the specified distributed job workers cannot communicate over required network path failure. The matching remediation is: Allow the narrowly scoped supported compute-to-compute communication paths and correct the distributed training rendezvous configuration.

B:

Write the model into the declared Azure ML job output path and track the versioned artifact used in registration. This could address training job does not persist output artifacts, but it does not solve the specified distributed job workers cannot communicate over required network path failure. The matching remediation is: Allow the narrowly scoped supported compute-to-compute communication paths and correct the distributed training rendezvous configuration.

D:

Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls. This could address overnight bulk scoring saturates interactive online endpoint, but it does not solve the specified distributed job workers cannot communicate over required network path failure. The matching remediation is: Allow the narrowly scoped supported compute-to-compute communication paths and correct the distributed training rendezvous configuration.

Question 29

During a Foundry generative AI quality and observability incident, the following condition occurs: A GenAIOps release pipeline writes evaluation scores to a report but deploys an agent even when groundedness falls below the required threshold. The fix must preserve evaluation validity and avoid unnecessary user-data collection. Which production-safe remediation should the engineer apply to this AI workflow?

  1. Instrument cost attribution across generation and embedding services, cache reusable embeddings and alert on unplanned spend.
  2. Include appropriate complete conversation histories and tool events in the supported agent evaluation workflow while preserving privacy protections.
  3. Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.
  4. Correct the test dataset mapping for query, context, response and reference fields and validate a sampled evaluation before the full run.

Explanation

Correct Answer

C:

Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances. This targets evaluation workflow fails to block degraded agent deployment. Automated evaluation provides protection only when results are connected to enforceable deployment decisions. Verify the failure through: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. Validate success with: An intentionally degraded candidate fails before production promotion and a passing candidate proceeds with an auditable evaluation receipt.

Incorrect Answers

A:

Instrument cost attribution across generation and embedding services, cache reusable embeddings and alert on unplanned spend. This could address token cost report omits embedding requests, but it does not solve the specified evaluation workflow fails to block degraded agent deployment failure. The matching remediation is: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.

B:

Include appropriate complete conversation histories and tool events in the supported agent evaluation workflow while preserving privacy protections. This could address multi-turn agent evaluation uses isolated last message only, but it does not solve the specified evaluation workflow fails to block degraded agent deployment failure. The matching remediation is: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.

D:

Correct the test dataset mapping for query, context, response and reference fields and validate a sampled evaluation before the full run. This could address genai evaluation dataset lacks expected answer column mapping, but it does not solve the specified evaluation workflow fails to block degraded agent deployment failure. The matching remediation is: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.

Question 30

While preparing a secure Foundry model or agent release, the following condition occurs: A high-volume assistant deployment repeatedly hits throughput saturation during business-hour spikes because its provisioned throughput units were selected from average rather than peak token load. The remediation must meet enterprise access and reliability requirements without a broad bypass. Which TWO complementary choices identify the issue and correct it? Select TWO.

  1. Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest.
  2. Right-size the supported provisioned capacity or apply controlled request pacing and overflow strategy within budget and region limits.
  3. Inspect input/output token volume, peak concurrent requests, provisioned throughput allocation, throttling codes and deployment latency percentiles.
  4. Replace broad management roles with the minimum supported inference access permission and remove redundant elevated grants.
  5. Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements.

Explanation

Correct Answers

B:

Right-size the supported provisioned capacity or apply controlled request pacing and overflow strategy within budget and region limits. This targets provisioned throughput capacity insufficient at peak traffic. Provisioned model throughput constrains admitted tokens and request demand; average daily request counts cannot establish a safe peak capacity allocation. Verify the failure through: Inspect input/output token volume, peak concurrent requests, provisioned throughput allocation, throttling codes and deployment latency percentiles. Validate success with: Peak requests meet approved latency and throttling targets under a representative load test while capacity costs remain documented.

C:

Inspect input/output token volume, peak concurrent requests, provisioned throughput allocation, throttling codes and deployment latency percentiles. This is the most relevant evidence for provisioned throughput capacity insufficient at peak traffic. Provisioned model throughput constrains admitted tokens and request demand; average daily request counts cannot establish a safe peak capacity allocation. The matching controlled remediation is: Right-size the supported provisioned capacity or apply controlled request pacing and overflow strategy within budget and region limits.

Incorrect Answers

A:

Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest. This is associated with managed foundation model deployment lacks compatible runtime image, not the essential evidence-and-fix pair for provisioned throughput capacity insufficient at peak traffic. First examine: Inspect input/output token volume, peak concurrent requests, provisioned throughput allocation, throttling codes and deployment latency percentiles. Then correct: Right-size the supported provisioned capacity or apply controlled request pacing and overflow strategy within budget and region limits.

D:

Replace broad management roles with the minimum supported inference access permission and remove redundant elevated grants. This is associated with foundry administrator role granted to consumer application, not the essential evidence-and-fix pair for provisioned throughput capacity insufficient at peak traffic. First examine: Inspect input/output token volume, peak concurrent requests, provisioned throughput allocation, throttling codes and deployment latency percentiles. Then correct: Right-size the supported provisioned capacity or apply controlled request pacing and overflow strategy within budget and region limits.

E:

Choose an appropriate available model and supported prompt chunking strategy that meets the workload's context and cost requirements. This is associated with foundation model selected without required context window, not the essential evidence-and-fix pair for provisioned throughput capacity insufficient at peak traffic. First examine: Inspect input/output token volume, peak concurrent requests, provisioned throughput allocation, throttling codes and deployment latency percentiles. Then correct: Right-size the supported provisioned capacity or apply controlled request pacing and overflow strategy within budget and region limits.

Operationalizing Machine Learning and Generative AI Solutions (AI-300) Practice Tests