SOL-C01 Question #311: Real Exam Question with Answer & Explanation

Question

A data scientist has developed a Streamlit application within a Snowflake Notebook to perform predictive analytics on customer churn. The application uses a pre-trained machine learning model stored as a Snowflake stage object. The model takes several customer features as input, which are stored in a Snowflake table called 'CUSTOMER FEATURES'. The data scientist needs to ensure that the model is loaded efficiently and that the inference is performed securely within the Snowflake environment, minimizing data movement. Which of the following approaches would be the MOST efficient and secure for loading the pre- trained model and performing the inference within the Snowflake environment using a Streamlit application?

Options

• ADownload the pre-trained model from the Snowflake stage to the Streamlit application's local
• BCreate a Snowflake external function (UDF) that loads the pre-trained model from the Snowflake
• CUse Streamlit's caching capabilities Cst.cache_resource') to load the pre-trained model from the
• DCreate a Snowflake stored procedure that loads the pre-trained model from the Snowflake stage
• EImplement Snowpark and load the model from the Snowflake stage, transforming and sending the

Explanation

Options B and E are the most appropriate for efficiency and security. Creating a Snowflake UDF or Snowpark session (E) keeps the data and model processing within the Snowflake environment, minimizing data movement and leveraging Snowflake's compute resources. UDFs or Snowpark session provide a secure and efficient way to perform the inference. Downloading model is not scalable (A), simple caching might be inefficent (C), storing procedure is not suitable as its not SQL based (D).

No community discussion yet for this question.