SnowPro Specialty: Gen AI · 10% of the exam

Document Processing: free practice questions

5 sample questions from our 25-question bank for this domain — answers and explanations included. These are the same scenario-based style as the real Snowflake exam.

1. A new team member proposes building a brand-new document pipeline using the original Document AI model-training UI they learned on a previous version of Snowflake. What should they be told?

  • A. That UI is retired; new pipelines should use the SQL-callable Cortex document-processing functions such as PARSE_DOCUMENT and AI_EXTRACT instead✓ Correct
  • B. The UI is still the recommended approach for all new pipelines
  • C. The UI has simply been renamed to AI_EXTRACT but otherwise works identically
  • D. The UI now requires requesting special access from Snowflake support before use
Explanation

Document AI's model-training UI has been retired; the current, exam-relevant surface is the SQL-callable Cortex functions that accomplish the same goal. It hasn't merely been renamed to AI_EXTRACT, and there's no support-access workaround to revive a retired UI.

2. You are assembling an end-to-end document-to-answer pipeline. Which ordering is correct?

  • A. Parse documents → chunk text → embed chunks → index/retrieve (Cortex Search or vector similarity) → COMPLETE with retrieved context✓ Correct
  • B. Embed first → then parse the PDF → then delete everything
  • C. COMPLETE first → parse later → never index
  • D. Translate → VACUUM → OPTIMIZE → done
Explanation

A document RAG pipeline parses files to text, chunks it, embeds the chunks, indexes/retrieves the relevant ones (Cortex Search or vector similarity), and passes them to COMPLETE as grounding context. The other orderings are nonsensical.

3. A team needs the raw text of scanned personal letters for keyword search only; no tables or headings in the source carry meaning. Which is the more appropriate, cost-effective choice?

  • A. OCR mode, since structure preservation isn't needed here✓ Correct
  • B. LAYOUT mode, since it should always be preferred regardless of content
  • C. Neither mode — call AI_EXTRACT directly on the raw file
  • D. Neither mode — call SPLIT_TEXT_RECURSIVE_CHARACTER directly on the raw file
Explanation

When structure carries no meaning, OCR mode's plain text extraction is sufficient and simpler. LAYOUT mode isn't wrong here, just unnecessary; AI_EXTRACT and chunking both require parsed text as input, not a raw file.

4. A pipeline must extract text from scanned PDFs stored on a stage so the content can be embedded for RAG. Which Cortex function performs OCR and layout-aware text extraction from documents?

  • A. PARSE_DOCUMENT (AI_PARSE_DOCUMENT)✓ Correct
  • B. SENTIMENT
  • C. VECTOR_COSINE_SIMILARITY
  • D. COPY INTO
Explanation

PARSE_DOCUMENT (AI_PARSE_DOCUMENT) extracts text from documents such as PDFs, including OCR and layout awareness, producing text you can then chunk and embed. The other functions score sentiment, compare vectors, and load data respectively.

5. Which of these correctly orders the stages of a document-to-RAG pipeline?

  • A. Parse the document, chunk the text, embed the chunks, index them, retrieve relevant ones, then ground the generation prompt✓ Correct
  • B. Embed the document, parse it, chunk it, retrieve chunks, then index them
  • C. Retrieve first, then parse, embed, and chunk in any order
  • D. Index the document, retrieve it, then parse and chunk it afterward
Explanation

Each stage depends on the output of the one before it, so parse → chunk → embed → index → retrieve → ground is the only order that works: there's nothing to chunk before parsing, nothing to embed before chunking, and nothing to retrieve before indexing.

20 more questions in this domain

Practice the full bank with instant grading, flashcards, and a timed mock exam.

Start practicing free