4.2 Preparing Data for AI and ML PROFESSIONAL-DATA-ENGINEER Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A customer support team is building a RAG system to answer queries from a knowledge base of troubleshooting guides. They have generated embeddings for each guide and stored them in a vector database. When a user asks a question, the system retrieves the top 5 most similar chunks and sends them to a language model for answer generation. The answers are often incomplete because the relevant information is spread across multiple chunks. What should they do to improve the quality of answers?

   • AReduce the number of retrieved chunks to 1 to avoid noise.
   • BUse a larger language model without changing retrieval.
   • CIncrease the number of retrieved chunks and use a re-ranking model to select the most relevant ones.
   • DGenerate a single embedding for the entire knowledge base.
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   When preparing data for a BigQuery ML model, which data type is required for a feature column that will be used as a categorical input?

   • AARRAY<STRING>
   • BFLOAT64
   • CSTRING
   • DBYTES
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   You are training a linear regression model in BigQuery ML. Which feature engineering technique is most appropriate to ensure that a feature with a range of 0–1000 and another with a range of 0–1 contribute equally to the model?

   • ABinning
   • BOne-hot encoding
   • CStandardization (z-score scaling)
   • DFeature selection
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   When storing embeddings for downstream use, which Google Cloud database is optimized for vector similarity search?

   • ACloud SQL
   • BCloud Spanner
   • CVertex AI Vector Search
   • DBigQuery
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A news organization is building a RAG system to answer questions about archived articles. They have a large corpus of articles in multiple languages. The team wants to ensure that retrieval works well across languages. They are considering using a multilingual embedding model. What is the most important consideration when preparing the data for this system?

   • AChunk articles consistently and store language metadata to enable filtering and potential re-ranking.
   • BUse separate embedding models for each language and store them in different vector stores.
   • CGenerate embeddings only for the most popular languages.
   • DTranslate all articles to English before embedding.
   Reveal Solution Discussion Select an answer first