Databricks Certified Generative AI Engineer Associate Bundle

Not the strongest safeguard choice for this specific question. Do not assert every index feature or hybrid/rerank option exists in every SKU, workspace or region. This action primarily relates to Standard versus storage-optimized Vector Search index: Index choice trades capacity, update speed, latency and storage economics; availability and feature support may vary by cloud/region. However, the described case — A production team passes around unversioned notebook checkpoints and cannot reproduce the active agent package. — is testing Unity Catalog registered model version and privileges. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Enforce reproducible environment manifests and API compatibility checks in CI/CD. This action primarily relates to Deployed model dependency drift and environment compatibility: Models logged in experiments can fail in serving if Python dependencies or inference signatures differ between build and target. However, the described case — A production team passes around unversioned notebook checkpoints and cannot reproduce the active agent package. — is testing Unity Catalog registered model version and privileges. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Treat index freshness as a monitored dependency of source-document correctness. This action primarily relates to Databricks Vector Search index from governed Delta data: A vector index serves embeddings of chunked records maintained in a Unity Catalog governed Delta source table. However, the described case — A production team passes around unversioned notebook checkpoints and cannot reproduce the active agent package. — is testing Unity Catalog registered model version and privileges. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Filter toxic or instruction-like content and use reviewed trusted source subsets for customer-facing retrieval. This action primarily relates to Toxic, inaccurate and adversarial corpus cleaning: Unreliable source text can teach unsafe behavior through retrieval; quarantine and source review matter before embeddings. However, the described case — A startup scrapes a paid technical manual site into its product RAG index despite restrictive subscription terms. — is testing Training and retrieval corpus license rights. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Apply moderation, require human review and restrict public-post tools to approved operators. This action primarily relates to Prompt safety filtering and tool allowlists: Content filters and action allowlists are complementary; a moderation score is not a substitute for authorization of external effects. However, the described case — A startup scrapes a paid technical manual site into its product RAG index despite restrictive subscription terms. — is testing Training and retrieval corpus license rights. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Isolate source text in retrieval context and use backend authorization for ticket actions. This action primarily relates to Input guardrails against prompt-injected source documents: A document that says 'ignore your rules' is untrusted data; a robust agent must enforce instruction and tool boundaries independently. However, the described case — A startup scrapes a paid technical manual site into its product RAG index despite restrictive subscription terms. — is testing Training and retrieval corpus license rights. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Grant only the tools required by the business workflow; general language understanding is not action authorization. This action primarily relates to Agent tool wrappers versus direct LLM responses: A tool-using agent needs explicit callable interfaces and typed tool results, whereas plain generation requires no action execution. However, the described case — A support assistant reveals hidden instructions when users ask it to ignore the normal helpdesk policy. — is testing Unsafe model output and jailbreak response diagnosis. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Never trust an LLM-extracted customer identifier without server-side permission validation. This action primarily relates to Query-field driven prompt and retrieval filters: User intent and structured fields such as account_id should drive authorized SQL lookups or retrieval filters, not be ignored in free text. However, the described case — A support assistant reveals hidden instructions when users ask it to ignore the normal helpdesk policy. — is testing Unsafe model output and jailbreak response diagnosis. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Document supported decoding options per endpoint and do not assume temperature guarantees factual correctness. This action primarily relates to Temperature and deterministic format tradeoffs: Decoding settings can affect variability and cost but cannot replace validation, grounding or application logic. However, the described case — A support assistant reveals hidden instructions when users ask it to ignore the normal helpdesk policy. — is testing Unsafe model output and jailbreak response diagnosis. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Select a suitable embedding model rather than treating generated prose as a numeric embedding. This action primarily relates to Specialized model versus general LLM: Embedding, reranking, extraction and generative models solve different stages and should be evaluated for their actual role. However, the described case — A ledger explanation app needs repeatable short outputs, but developers increase temperature to make wording more creative. — is testing Temperature and deterministic format tradeoffs. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Revise few-shot examples for format compliance and rerun previously passing examples. This action primarily relates to Prompt changes evaluated against baseline: A prompt improvement should be measured against a fixed baseline set, not accepted because a few outputs sound better. However, the described case — A ledger explanation app needs repeatable short outputs, but developers increase temperature to make wording more creative. — is testing Temperature and deterministic format tradeoffs. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Pin the retrieval index version and trace reference set in experiments before comparing runs. This action primarily relates to Experiment tracking and version reproducibility: MLflow experiments should retain prompt, model, dataset, parameters and evaluation outputs to support a defensible comparison. However, the described case — A ledger explanation app needs repeatable short outputs, but developers increase temperature to make wording more creative. — is testing Temperature and deterministic format tradeoffs. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Inspect each source connector's contract, join keys, runtime error handling and the final prompt inputs. This action primarily relates to Parallel retrieval and deterministic chain postprocessing: Combining structured lookup and semantic retrieval requires explicit routing, result joining, and downstream schema checks. However, the described case — A benefits intake app requires a JSON object with claim_id, decision and reason but the assistant returns Markdown paragraphs. — is testing Structured JSON schema prompt and format compliance. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Determine whether output is fixed fields, document answers or orchestration of multiple tasks with tool calls. This action primarily relates to Agent Bricks Information Extraction versus Multiagent Supervisor: Information Extraction maps source documents to structured fields; Multiagent Supervisor coordinates specialized capabilities for compound workflows. However, the described case — A benefits intake app requires a JSON object with claim_id, decision and reason but the assistant returns Markdown paragraphs. — is testing Structured JSON schema prompt and format compliance. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Trace which chain stage receives the question, retriever matches, prompt context and generated answer. This action primarily relates to Retriever-generator chain component boundaries: A basic RAG chain separates user input, query transformation, retrieval, prompt assembly, model generation and output validation. However, the described case — A benefits intake app requires a JSON object with claim_id, decision and reason but the assistant returns Markdown paragraphs. — is testing Structured JSON schema prompt and format compliance. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Use task-appropriate sampling and assess diversity and usefulness rather than exact wording alone. This action primarily relates to Temperature and deterministic format tradeoffs: Decoding settings can affect variability and cost but cannot replace validation, grounding or application logic. However, the described case — A compliance summary app adds examples but stops honoring the one-line required answer format. — is testing Prompt changes evaluated against baseline. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Log model/prompt/data identifiers and metrics to a run with reproducible evaluation artifacts. This action primarily relates to Experiment tracking and version reproducibility: MLflow experiments should retain prompt, model, dataset, parameters and evaluation outputs to support a defensible comparison. However, the described case — A compliance summary app adds examples but stops honoring the one-line required answer format. — is testing Prompt changes evaluated against baseline. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Require preproduction evaluation on curated labeled examples and later observe production quality. This action primarily relates to Offline evaluation versus ongoing production monitoring: Offline test datasets catch regressions pre-release; production monitoring detects drift, real-user failures and evolving prompts. However, the described case — A compliance summary app adds examples but stops honoring the one-line required answer format. — is testing Prompt changes evaluated against baseline. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Apply supported AI Gateway endpoint quotas or rate limits by workload and inspect usage traces. This action primarily relates to AI Gateway rate limits, usage tables and inference telemetry: AI Gateway offers controlled access, supported usage tracking and rate limiting to protect shared endpoints and measure consumption. However, the described case — An invoice extraction agent passes generic relevance judges but loses an essential invoice currency field. — is testing MLflow GenAI tracing, built-in and custom scorers. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Define scoring anchors, calibrate experts on disputed samples and use aligned feedback as reliable evaluation labels. This action primarily relates to Human SME feedback rubric and reviewer calibration: Domain experts should use a shared rubric, calibrate disagreements and maintain adjudicated examples to improve model evaluation. However, the described case — An invoice extraction agent passes generic relevance judges but loses an essential invoice currency field. — is testing MLflow GenAI tracing, built-in and custom scorers. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest implement choice for this specific question. Choose the candidate meeting the agreed citation-quality threshold and p95 latency with lower cost. This action primarily relates to Model selection from quantitative task metrics: Select candidate LLMs by measured task success, latency, safety and token cost, not parameter count or leaderboard rank. However, the described case — An invoice extraction agent passes generic relevance judges but loses an essential invoice currency field. — is testing MLflow GenAI tracing, built-in and custom scorers. It does not directly establish the implement outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Inspect whether the task is grounded document Q&A, extraction, tool coordination or cross-agent orchestration. This action primarily relates to Agent Bricks Knowledge Assistant suitability: A governed knowledge assistant grounds conversational responses in curated documents and access-controlled retrieval. However, the described case — A finance controller wants a one-paragraph explanation of a long audit memo, not a risk score. — is testing Task choice: classification versus generative summarization. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Inspect the business decision, data owners, actor permissions, known inputs and success criteria. This action primarily relates to Translate use-case requirements into pipeline inputs and outputs: Business acceptance criteria must map to concrete source inputs, generated fields, correctness measures and permission constraints. However, the described case — A finance controller wants a one-paragraph explanation of a long audit memo, not a risk score. — is testing Task choice: classification versus generative summarization. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Trace which chain stage receives the question, retriever matches, prompt context and generated answer. This action primarily relates to Retriever-generator chain component boundaries: A basic RAG chain separates user input, query transformation, retrieval, prompt assembly, model generation and output validation. However, the described case — A finance controller wants a one-paragraph explanation of a long audit memo, not a risk score. — is testing Task choice: classification versus generative summarization. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Inspect source origin, prompt hierarchy, tool permissions and trace evidence of instruction-following attempts. This action primarily relates to Input guardrails against prompt-injected source documents: A document that says 'ignore your rules' is untrusted data; a robust agent must enforce instruction and tool boundaries independently. However, the described case — A customer-support RAG source includes cardholder names and full card numbers even though questions concern card dispute policy. — is testing Sensitive-field masking before embedding. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Inspect filter decisions, action permissions, identity propagation and resulting audit events. This action primarily relates to Prompt safety filtering and tool allowlists: Content filters and action allowlists are complementary; a moderation score is not a substitute for authorization of external effects. However, the described case — A customer-support RAG source includes cardholder names and full card numbers even though questions concern card dispute policy. — is testing Sensitive-field masking before embedding. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest diagnose choice for this specific question. Inspect which fields are necessary, masking false positives and any leaked sensitive fields. This action primarily relates to Contextual masking versus destructive data removal: Masking must preserve information needed for the task, such as non-sensitive dates or amounts, and cannot replace access control. However, the described case — A customer-support RAG source includes cardholder names and full card numbers even though questions concern card dispute policy. — is testing Sensitive-field masking before embedding. It does not directly establish the diagnose outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest verify choice for this specific question. Simulate quality and latency threshold breaches and confirm responsible owners receive alerts. This action primarily relates to Production performance KPIs and incident thresholds: Monitoring should include answer correctness, retrieval coverage, safety, latency, drift and per-user harm indicators, not just availability. However, the described case — A finance team cannot attribute escalating LLM spend to teams because requests have no workload identity. — is testing AI Gateway rate limits, usage tables and inference telemetry. It does not directly establish the verify outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest verify choice for this specific question. Test a missing-ground-truth record to verify the judge abstains or uses the appropriate alternate scorer. This action primarily relates to Ground-truth requirement for correctness versus reference-free judges: A correctness comparison often needs an expected answer; other judges such as format or safety can operate without a reference. However, the described case — A finance team cannot attribute escalating LLM spend to teams because requests have no workload identity. — is testing AI Gateway rate limits, usage tables and inference telemetry. It does not directly establish the verify outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest verify choice for this specific question. Re-evaluate previously disputed cases and confirm inter-rater agreement improves after calibration. This action primarily relates to Human SME feedback rubric and reviewer calibration: Domain experts should use a shared rubric, calibrate disagreements and maintain adjudicated examples to improve model evaluation. However, the described case — A finance team cannot attribute escalating LLM spend to teams because requests have no workload identity. — is testing AI Gateway rate limits, usage tables and inference telemetry. It does not directly establish the verify outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Retest adversarial sources after adding new file types and tool connectors. This action primarily relates to Input guardrails against prompt-injected source documents: A document that says 'ignore your rules' is untrusted data; a robust agent must enforce instruction and tool boundaries independently. However, the described case — A chatbot accepts a malicious request to publish harassing content to a company social feed. — is testing Prompt safety filtering and tool allowlists. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Use layered permissions and field-level handling instead of broad redaction that destroys task utility. This action primarily relates to Contextual masking versus destructive data removal: Masking must preserve information needed for the task, such as non-sensitive dates or amounts, and cannot replace access control. However, the described case — A chatbot accepts a malicious request to publish harassing content to a company social feed. — is testing Prompt safety filtering and tool allowlists. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.

Not the strongest safeguard choice for this specific question. Keep masking rules under regression control and treat embeddings and logs as potentially sensitive too. This action primarily relates to Sensitive-field masking before embedding: Masking or tokenization can remove unnecessary PII from chunks before embedding and external model calls while retaining useful meaning. However, the described case — A chatbot accepts a malicious request to publish harassing content to a company social feed. — is testing Prompt safety filtering and tool allowlists. It does not directly establish the safeguard outcome or constraint that the question requires; it may still be useful in a different stage or incident.