Which activity is part of managing the lifecycle of an external AI model?

Monitoring model performance and retraining when necessary

Defining the model's input schema

Creating the model's training dataset

A research organization stores lengthy scientific papers in a database. They want to generate embeddings for semantic search. The embedding model has a token limit of 512 tokens. The papers contain sections, subsections, and paragraphs. What is the best practice for chunking these documents?

Generate one embedding per paper using a summarization model first.

Use the entire paper as a single chunk, truncating to 512 tokens.

Chunk by fixed character count (e.g., 1000 characters) regardless of content structure.

Chunk by semantic boundaries (e.g., sections or paragraphs) and keep each chunk under 512 tokens.

A financial services company wants to integrate an external AI model into their Azure SQL database to extract structured data (e.g., account numbers and dates) from unstructured text. They need the model to return output in a strict JSON schema. Which capability is most important when evaluating external models?

Larger model size for better general knowledge

Structured output support (e.g., JSON mode or function calling)

Multimodal capabilities (e.g., image and text processing)

Microsoft Certified:SQL AI Developer Associate

Domain 3Objective 1

Design and Implement Models and Embeddings DP-800 Practice Questions (Page 6)

Part of the Implement AI capabilities in database solutions domain, which accounts for 25–30% of the DP-800 exam. Microsoft does not publish an official question count, but from its 120-minute exam (~50–80 total, ~13–24 in this domain), expect 4–8 from this objective — we provide 33 practice questions to prepare you well beyond it. (estimate)

33questions here

7free pages

7concepts

25–30%of the exam

Questions 26–30

26foundation · easy
Which activity is part of managing the lifecycle of an external AI model?
Select an answer first
27foundation · easy
Which method for maintaining embeddings uses database triggers to automatically update embeddings when source data changes?
Select an answer first
28application · medium
A research organization stores lengthy scientific papers in a database. They want to generate embeddings for semantic search. The embedding model has a token limit of 512 tokens. The papers contain sections, subsections, and paragraphs. What is the best practice for chunking these documents?
Select an answer first
29foundation · easy
Which Azure service is commonly used to generate embeddings from text data for integration into a database solution?
Select an answer first
30application · medium
A financial services company wants to integrate an external AI model into their Azure SQL database to extract structured data (e.g., account numbers and dates) from unstructured text. They need the model to return output in a strict JSON schema. Which capability is most important when evaluating external models?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to DP-800

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Microsoft. “DP-800” is a trademark of its owner, used for identification only.

Design and Implement Models and Embeddings DP-800 Practice Questions (Page 6)

Questions 26–30

Which activity is part of managing the lifecycle of an external AI model?

Which method for maintaining embeddings uses database triggers to automatically update embeddings when source data changes?

A research organization stores lengthy scientific papers in a database. They want to generate embeddings for semantic search. The embedding model has a token limit of 512 tokens. The papers contain sections, subsections, and paragraphs. What is the best practice for chunking these documents?

Which Azure service is commonly used to generate embeddings from text data for integration into a database solution?

Keep studying — Implement AI capabilities in database solutions