McAfee Secure

Exam Code: DP-800

Exam Name: Developing AI-Enabled Database Solutions

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: SQL AI Developer Associate

Microsoft DP-800 Questions & Answers

Study with Up-To-Date REAL Exam Questions and Answers from the ACTUAL Test

101 Questions & Answers with Testing Engine

"Developing AI-Enabled Database Solutions Exam", also known as DP-800 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date DP-800 Exam Engine. Your DP-800 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft DP-800 Practice Test 34

Question 330

A finance application needs deterministic machine-readable response fields from model inference. What should the SQL developer do FIRST to address the specific issue?

  1. Benchmark smaller models and route complex cases only when quality needs justify it.
  2. Prefer a model and API mode supporting validated structured output and enforce schema checks.
  3. Plan a new compatible vector column/index and re-embed the corpus using the new model.
  4. Include semantically meaningful title and content fields, with identifiers retained as metadata.

Explanation

Correct Answer

B:

Prefer a model and API mode supporting validated structured output and enforce schema checks. This targets the primary issue because unconstrained prose is fragile to parse and can omit required keys. Verification: Sample calls produce valid JSON schema instances or explicit errors. Prevent recurrence: Reject unexpected fields and never trust generated totals without validation.

Incorrect Answers

A:

Model size increases cost and latency without guaranteeing proportional quality gains.

C:

Vector dimensionality is fixed by the column and must match the query vector.

D:

IDs alone lack the natural-language signal needed for semantic retrieval.

Question 331

A similarity API submits a query vector of different dimension than stored vectors. Which acceptance check is MOST directly related to this scenario?

  1. Known nearest-neighbor fixtures appear first in the final order.
  2. Known phrase and inflection tests return expected articles.
  3. The corrected query executes and returns expected similar records.
  4. The quoted phrase test excludes documents without that phrase.

Explanation

Correct Answer

C:

The corrected query executes and returns expected similar records. This is the decisive observable evidence after implementing generate query embeddings with the same approved model and dimension as the column. It tests the underlying requirement: Vector operations require dimension compatibility.

Incorrect Answers

A:

Known nearest-neighbor fixtures appear first in the final order. However, that result would not establish the required outcome: the corrected query executes and returns expected similar records.

B:

Known phrase and inflection tests return expected articles. However, that result would not establish the required outcome: the corrected query executes and returns expected similar records.

D:

The quoted phrase test excludes documents without that phrase. However, that result would not establish the required outcome: the corrected query executes and returns expected similar records.

Question 332

A nonclustered index becomes misaligned with a heavily partitioned fact table, complicating switches. Which technical property or constraint MOST directly matters here?

  1. Metadata-based partition switching requires strict schema and partition alignment.
  2. A surrogate primary key alone does not protect a separate business identifier.
  3. Referential integrity prevents orphan children unless cascade or other actions are deliberately configured.
  4. Misaligned indexes can block efficient partition maintenance.

Explanation

Correct Answer

D:

Misaligned indexes can block efficient partition maintenance. This explains why the corrective action is to align the relevant indexes and partition scheme where switching operations require it. Confirm the diagnosis when a maintenance rehearsal successfully switches a partition with required constraints.

Incorrect Answers

A:

Metadata-based partition switching requires strict schema and partition alignment. The actual technical constraint is that misaligned indexes can block efficient partition maintenance.

B:

A surrogate primary key alone does not protect a separate business identifier. The actual technical constraint is that misaligned indexes can block efficient partition maintenance.

C:

Referential integrity prevents orphan children unless cascade or other actions are deliberately configured. The actual technical constraint is that misaligned indexes can block efficient partition maintenance.

Question 333

A trigger sends an external email during every sale and delays checkouts. Which preventive practice BEST addresses the underlying risk over time?

  1. Implement transactional outbox or supported change notifications.
  2. Minimize trigger side effects and audit dependent trigger paths.
  3. Restrict callers to authorized tenant and customer context.
  4. Implement idempotent retry and bounded transaction duration.

Explanation

Correct Answer

A:

Implement transactional outbox or supported change notifications. This guards against recurrence because synchronous trigger work participates in the calling transaction and can increase latency. The direct remediation is: Move external notifications to a reliable asynchronous event workflow after durable commit.

Incorrect Answers

B:

Minimize trigger side effects and audit dependent trigger paths. The current task instead requires this action: implement transactional outbox or supported change notifications.

C:

Restrict callers to authorized tenant and customer context. The current task instead requires this action: implement transactional outbox or supported change notifications.

D:

Implement idempotent retry and bounded transaction duration. The current task instead requires this action: implement transactional outbox or supported change notifications.

Question 334

An endpoint returns a 503 error body and the procedure attempts to parse it as a successful model completion. What should the SQL developer do FIRST to address the specific issue?

  1. Parse the supported response schema and extract the correct generated content field.
  2. Check HTTP status and error fields before extracting generated output.
  3. Use deterministic SQL aggregation for arithmetic and add RAG only for explanatory source text if needed.
  4. Use tenant-trimmed retrieval with citations and a strict abstention policy.

Explanation

Correct Answer

B:

Check HTTP status and error fields before extracting generated output. This targets the primary issue because transport errors must be separated from valid inference responses. Verification: A simulated 503 is routed to retry/alert logic without false answer content. Prevent recurrence: Log safe diagnostics and maintain circuit-breaker policy.

Incorrect Answers

A:

HTTP response metadata and generated text are different parts of the payload.

C:

Numeric reconciliation is a structured calculation, not a reason to hallucinate an answer from documents.

D:

RAG is appropriate only if the retriever enforces contractual permissions.

Question 335

An application calls a procedure without qualifying schema names, yielding inconsistent execution context. Which acceptance check is MOST directly related to this scenario?

  1. Attack strings are treated as data and queries return only intended rows.
  2. A test UPDATE completes once with expected values and no recursion error.
  3. Tests resolve the intended procedure and handle null/invalid inputs.
  4. Boundary amounts return correct DECIMAL values in unit tests.

Explanation

Correct Answer

C:

Tests resolve the intended procedure and handle null/invalid inputs. This is the decisive observable evidence after implementing use schema-qualified procedure references and explicit input/output parameter contracts. It tests the underlying requirement: Unclear resolution and parameter types create ambiguous deployment behavior.

Incorrect Answers

A:

Attack strings are treated as data and queries return only intended rows. However, that result would not establish the required outcome: tests resolve the intended procedure and handle null/invalid inputs.

B:

A test UPDATE completes once with expected values and no recursion error. However, that result would not establish the required outcome: tests resolve the intended procedure and handle null/invalid inputs.

D:

Boundary amounts return correct DECIMAL values in unit tests. However, that result would not establish the required outcome: tests resolve the intended procedure and handle null/invalid inputs.

Question 336

A model call from SQL fails with HTTP 401 despite a correct JSON body. Which technical principle BEST explains the relevant design decision?

  1. The stored procedure enables outbound HTTP but requires correct permissions, endpoint reachability and payload formatting.
  2. Sending unauthorized rows to a model is a disclosure even if its final answer omits them.
  3. RAG can inject up-to-date source evidence at answer time rather than encoding it in weights.
  4. An authorization error occurs before model inference can produce valid content.

Explanation

Correct Answer

D:

An authorization error occurs before model inference can produce valid content. This explains why the corrective action is to check REST endpoint authentication and configured identity before changing the SQL result parser. Confirm the diagnosis when the approved identity receives a 2xx response while an unauthorized one fails.

Incorrect Answers

A:

The stored procedure enables outbound HTTP but requires correct permissions, endpoint reachability and payload formatting. The actual technical constraint is that an authorization error occurs before model inference can produce valid content.

B:

Sending unauthorized rows to a model is a disclosure even if its final answer omits them. The actual technical constraint is that an authorization error occurs before model inference can produce valid content.

C:

RAG can inject up-to-date source evidence at answer time rather than encoding it in weights. The actual technical constraint is that an authorization error occurs before model inference can produce valid content.

Question 337

A payroll database must keep salary values unreadable even to privileged server-side SQL operators. Following remediation, what safeguard BEST prevents a repeat of this specific failure?

  1. Protect client-side keys and plan query limitations and rotation.
  2. Separate key custodianship and audit key usage.
  3. Reset pooled connections and validate context on each request.
  4. Monitor service-principal grants and run negative authorization tests.

Explanation

Correct Answer

A:

Protect client-side keys and plan query limitations and rotation. This guards against recurrence because always Encrypted protects configured column values from the SQL engine during normal processing. The direct remediation is: Use Always Encrypted with keys managed outside the database and supported client drivers.

Incorrect Answers

B:

Separate key custodianship and audit key usage. The current task instead requires this action: protect client-side keys and plan query limitations and rotation.

C:

Reset pooled connections and validate context on each request. The current task instead requires this action: protect client-side keys and plan query limitations and rotation.

D:

Monitor service-principal grants and run negative authorization tests. The current task instead requires this action: protect client-side keys and plan query limitations and rotation.

Question 338

A stored procedure sends query results to an LLM before validating tenant-level row access. Which technical action BEST resolves the primary problem?

  1. Summarize or chunk results within supported token and request limits before model invocation.
  2. Apply database RLS and data minimization before constructing the outbound payload.
  3. Use tenant-trimmed retrieval with citations and a strict abstention policy.
  4. Reconcile generated claims with SQL source values before publishing results.

Explanation

Correct Answer

B:

Apply database RLS and data minimization before constructing the outbound payload. This targets the primary issue because sending unauthorized rows to a model is a disclosure even if its final answer omits them. Verification: Captured requests from one tenant contain only that tenant's permitted data. Prevent recurrence: Audit outbound data scope and access context.

Incorrect Answers

A:

Summarize or chunk results within supported token and request limits before model invocation. The relevant technical condition instead is that apply database RLS and data minimization before constructing the outbound payload.

C:

RAG is appropriate only if the retriever enforces contractual permissions.

D:

Reconciling generated claims with SQL source values before publishing results addresses a different database behavior; the relevant distinction is that apply database RLS and data minimization before constructing the outbound payload.

Question 339

An internal Data API builder REST endpoint exposes every row anonymously. Which test result BEST demonstrates that the intended solution is working?

  1. A forged admin header fails while valid signed identities receive correct rights.
  2. Approved procedures decrypt the target values and unauthorized roles cannot open keys.
  3. An anonymous request is denied and an approved role sees only permitted fields.
  4. A simulated low-privilege role cannot infer restricted personal identifiers.

Explanation

Correct Answer

C:

An anonymous request is denied and an approved role sees only permitted fields. This is the decisive observable evidence after implementing configure authentication and per-entity role/operation permissions rather than public wildcard actions. It tests the underlying requirement: API generation does not automatically imply that exposed data is appropriately authorized.

Incorrect Answers

A:

A forged admin header fails while valid signed identities receive correct rights. However, that result would not establish the required outcome: an anonymous request is denied and an approved role sees only permitted fields.

B:

Approved procedures decrypt the target values and unauthorized roles cannot open keys. However, that result would not establish the required outcome: an anonymous request is denied and an approved role sees only permitted fields.

D:

A simulated low-privilege role cannot infer restricted personal identifiers. However, that result would not establish the required outcome: an anonymous request is denied and an approved role sees only permitted fields.

Question 340

Different reporting tools duplicate a sensitive sales join and occasionally disagree on calculations. Which technical principle BEST explains the relevant design decision?

  1. SCHEMABINDING protects referenced definitions against incompatible schema changes.
  2. SQL triggers run per statement and inserted can contain multiple rows.
  3. Nested or recursive trigger behavior can cause cycles and unexpected work.
  4. A view centralizes a query contract without storing a second copy by default.

Explanation

Correct Answer

D:

A view centralizes a query contract without storing a second copy by default. This explains why the corrective action is to create a governed view projecting the approved join and columns. Confirm the diagnosis when both tools return the same verified sales metric from the view.

Incorrect Answers

A:

SCHEMABINDING protects referenced definitions against incompatible schema changes. The actual technical constraint is that a view centralizes a query contract without storing a second copy by default.

B:

SQL triggers run per statement and inserted can contain multiple rows. The actual technical constraint is that a view centralizes a query contract without storing a second copy by default.

C:

Nested or recursive trigger behavior can cause cycles and unexpected work. The actual technical constraint is that a view centralizes a query contract without storing a second copy by default.

Developing AI-Enabled Database Solutions (DP-800) Practice Tests