Platform Services — Cortex AI, Iceberg, Notebooks, Git — Snowflake SnowPro Core practice questions

Materialized view vs Dynamic Table?

1. Two different names for the same auto-refreshed object
2. MVs pre-compute one query; DTs refresh to a target lag
3. Materialized views are deprecated in favor of dynamic tables
4. Dynamic Tables can only read data from external stages

Answer: B — MVs pre-compute one query; DTs refresh to a target lag

B) Correct — a materialized view is an auto-maintained result of a single query, while a dynamic table is declaratively refreshed to a target lag and can express multi-step pipelines. A) They are distinct objects. C) MVs remain supported. D) Dynamic tables read Snowflake tables.

Which statement about Apache Iceberg tables in Snowflake is TRUE?

1. They must always be managed by an external catalog such as AWS Glue
2. They convert Parquet files into Snowflake's proprietary micro-partition format on load
3. They store data in open formats in customer-managed cloud storage via an external volume
4. They cannot be queried alongside native Snowflake tables in the same statement

Answer: C — They store data in open formats in customer-managed cloud storage via an external volume

A) Snowflake can act as the Iceberg catalog itself; an external catalog is optional. B) Iceberg tables keep data in open Parquet/Iceberg format — no conversion to micro-partitions. C) Correct — data lives in the customer's storage referenced through an external volume. D) Iceberg and native tables can be joined in one query.

When Snowflake Copilot generates a SQL suggestion, what is the user's responsibility BEFORE relying on the results?

1. Nothing; Copilot output is covered by a correctness SLA
2. Grant Copilot the ACCOUNTADMIN role so it can validate the query
3. Review and run the SQL, checking it answers the question
4. Send the query to an external LLM for a second opinion

Answer: C — Review and run the SQL, checking it answers the question

C) Correct — the user must inspect, execute, and validate the suggested SQL before trusting its results. A) LLM suggestions carry no correctness guarantee. B) Copilot works within the user's existing privileges, and elevating them is unsafe. D) Sending the query outside is neither required nor governed.

Snowpark Container Services lets you run:

1. Python scripts submitted as stored procedures only
2. SQL statements scheduled through serverless tasks
3. Java user-defined functions packaged as JAR files
4. Long-lived containerized apps on Snowflake compute

Answer: D — Long-lived containerized apps on Snowflake compute

D) Correct — Snowpark Container Services runs OCI container images, including Streamlit apps, vector databases and batch jobs, on Snowflake-managed compute pools. A) and C) Describe Snowpark UDFs and procedures. B) Describes tasks, a different scheduling feature.

Cortex SUMMARIZE() function is invoked via:

1. A Python client library installed on the calling host
2. A REST endpoint called with an OAuth bearer token only
3. Standard SQL: SELECT SNOWFLAKE.CORTEX.SUMMARIZE(col)
4. The Snowpark session API and no other interface

Answer: C — Standard SQL: SELECT SNOWFLAKE.CORTEX.SUMMARIZE(col)

C) Correct — Cortex LLM functions are first-class SQL functions in the SNOWFLAKE.CORTEX schema. A) and D) Python and Snowpark access exist but are not the primary interface. B) REST is not the only way to reach the function.

Snowflake Cortex provides:

1. Hosting of long-lived containers on managed compute pools
2. SQL-callable LLM functions plus Document AI features
3. Cloud object storage for unstructured data files
4. Authentication and federated identity management

Answer: B — SQL-callable LLM functions plus Document AI features

B) Correct — Cortex exposes serverless model functions such as SUMMARIZE, TRANSLATE, EXTRACT_ANSWER and EMBED_TEXT directly from SQL, alongside Document AI and fine-tuning. A) Containers are Snowpark Container Services. C) and D) Belong to other parts of the platform.

How are serverless Cortex LLM function calls billed?

1. Free of charge on any Enterprise edition Snowflake account
2. Per virtual warehouse second for the whole time the function runs
3. A flat monthly subscription charged for each licensed user
4. Credits consumed by usage, such as tokens the model processes

Answer: D — Credits consumed by usage, such as tokens the model processes

D) Correct — serverless Cortex LLM functions draw credits in proportion to usage, measured in tokens processed. A) Cortex is not bundled free with an edition. B) Serverless Cortex compute is billed separately from the calling warehouse. C) Snowflake does not sell Cortex per user.

Which capability lets business users ask questions of structured data in natural language, using a semantic model to map business terms to tables and columns?

1. Snowpipe Streaming
2. Dynamic tables
3. Cortex Analyst
4. Zero-copy cloning

Answer: C — Cortex Analyst

A) Snowpipe Streaming is a row-based ingestion API. B) Dynamic tables automate transformation pipelines, not natural-language querying. C) Correct — Cortex Analyst answers natural-language questions over structured data via a semantic model. D) Cloning creates instant object copies; it has no query interface.

Dynamic Tables are designed to replace which legacy pattern?

1. Standard views defined over one or more base tables
2. External tables pointing at files in cloud storage
3. Materialized views automatically maintained by Snowflake
4. Streams and Tasks chained into hand-built pipelines

Answer: D — Streams and Tasks chained into hand-built pipelines

D) Correct — a dynamic table declares its result and a target lag, so Snowflake handles the incremental refresh that previously required chained streams and tasks. A) Views are unaffected. B) External tables address a different need. C) Materialized views remain for narrower cases.

A data analyst wants to summarize support tickets stored in a table using SQL, without provisioning any ML infrastructure. Which Snowflake feature is the BEST fit?

1. Cortex AI LLM functions such as SUMMARIZE, called from SQL
2. Snowpark Container Services hosting a self-managed model server
3. External functions calling a cloud vendor's hosted ML API
4. Materialized views precomputing aggregates over the tickets

Answer: A — Cortex AI LLM functions such as SUMMARIZE, called from SQL

A) Correct — Cortex LLM functions run serverlessly and are invoked directly in SQL with no infrastructure to provision. B) Containers work but require managing images and compute pools. C) External functions add network configuration and an outside dependency. D) Materialized views precompute query results and cannot summarize text.

What is Snowpark, and what problem do Snowflake Notebooks solve?

Snowpark is a set of client APIs and runtimes (Python, Java, Scala) that let developers write DataFrame-style code and UDFs/procedures that execute inside Snowflake, so processing happens next to the data. Snowflake Notebooks provide a governed in-platform cell-based interface for exploring data and building pipelines and ML work in SQL and Python.

Name the SQL-callable Snowflake Cortex LLM functions and describe what they let a data team do without leaving SQL.

COMPLETE, SENTIMENT, SUMMARIZE, TRANSLATE and EXTRACT_ANSWER. They let analysts run generation, sentiment scoring, summarization, translation and question answering over governed Snowflake data directly in SQL, with no separate model hosting or data movement.

Conceptually, why do Snowflake Iceberg tables exist and how do they differ from native Snowflake tables?

Iceberg tables store data in open Parquet files with open Iceberg metadata in customer-owned cloud storage, so other engines can read the same data without copying. Native tables use Snowflake's proprietary micro-partition format in Snowflake-managed storage. Iceberg trades some proprietary optimization and feature coverage for interoperability and storage ownership.

Snowpark Container Services?

Run containerized apps (Streamlit, NeMo, vector DBs) directly on Snowflake compute.