Databricks Generative AI Engineer Associate Question 238
Select 2You are building a Retrieval-Augmented Generation (RAG) application using Databricks. The application needs to serve features such as document embeddings, vector similarity search, and real-time response generation. Which resources are essential to serve these features effectively?
- A
A vector database for storing and querying document embeddings.
- B
A GPU-enabled compute cluster for fine-tuning the language model in real-time.
- C
A feature store for managing real-time feature engineering pipelines.
- D
A language model hosted on an endpoint for generating responses.
- E
A batch processing system for periodic updates to the knowledge base.
Show answer and explanation
Correct answers: A, D
Explanation
To serve features for a RAG application, a vector database is required for storing and querying embeddings, and a hosted language model endpoint is necessary for generating responses in real time. Other resources, such as GPU-enabled compute or a feature store, are either not directly relevant or are more suited for tasks like model training or feature engineering.
- A. Correct.
A vector database is critical for storing and performing similarity searches on document embeddings, which is a key component of a RAG-based application.
- B. Incorrect.
While GPU-enabled compute may be useful during model training, it is not necessary for serving features like document embeddings or generating responses in real time.
- C. Incorrect.
A feature store is typically used for managing features in machine learning pipelines, but it is not directly relevant to serving document embeddings or response generation in a RAG application.
- D. Correct.
A hosted language model endpoint is essential for generating responses in real time, which is a core feature of a RAG-based application.
- E. Incorrect.
Batch processing for periodic updates may be useful for refreshing the knowledge base but is not critical for serving features in real time.