COF-C03 Question 119
Select 21.6 Explain AI/ML and application development featuresA retail company wants to build a customer-facing forecasting application directly in Snowflake. The development team needs to: (1) create and version application code in the data platform, (2) build simple machine learning predictions without moving data out of Snowflake, and (3) share the finished application with external customers as a Snowflake-native offering. Which combination of Snowflake features best meets these requirements?
- A
Use Snowpark to develop application logic, use Snowflake ML functions/models for in-platform ML workflows, and package/distribute the solution as a Native App Framework application
- B
Use external functions as the primary way to store and version application code, train models outside Snowflake only, and distribute the application through secure data sharing
- C
Use Git repository integration to manage source code in Snowflake, use Snowpark/Snowflake ML for in-platform development and ML, and distribute the application with the Native App Framework
- D
Use Streams and Tasks to build the customer-facing UI, use Cortex Search to train forecasting models, and distribute the application by sharing a virtual warehouse
Show answer and explanation
Correct answers: A, C
Explanation
The best answer is the combination of options 1 and 3 because they align with Snowflake's application development and AI/ML capabilities covered at the SnowPro Core level. Snowpark is Snowflake's developer framework for building data applications in-platform. Snowflake ML capabilities support machine learning workflows directly in Snowflake, which helps avoid unnecessary data movement. Git repository integration supports managing code as part of modern development practices. For distributing a customer-facing Snowflake-native solution, the Native App Framework is the intended feature, rather than secure data sharing alone. In contrast, external functions are for invoking external services, Streams and Tasks are operational/orchestration features, and Cortex Search is for search/retrieval scenarios rather than model training. These distinctions are consistent with Snowflake documentation for Snowpark, Snowflake ML, Git integration, and the Native App Framework.
- A. Correct.
Correct. Snowpark is the primary development framework for building applications in languages such as Python, Java, and Scala in Snowflake. Snowflake supports AI/ML development inside the platform through Snowflake ML capabilities, allowing teams to work with data without moving it externally. The Native App Framework is the correct mechanism for packaging and sharing a Snowflake-native application with consumers. This option accurately matches all three requirements, although code versioning is implied through platform development practices rather than explicitly called out.
- B. Incorrect.
Incorrect. External functions let Snowflake call out to services hosted outside Snowflake, but they are not a source-code storage or versioning mechanism. Requiring models to be trained outside Snowflake conflicts with the requirement to avoid moving data out of Snowflake. Secure data sharing is for sharing data, not packaging and distributing a full Snowflake-native application experience to customers in the way the Native App Framework is designed to do.
- C. Correct.
Correct. Git repository integration supports managing and synchronizing source code with Snowflake development workflows, satisfying the code creation/versioning requirement. Snowpark and Snowflake ML support application development and ML directly in Snowflake. The Native App Framework is specifically intended to build, package, and distribute Snowflake-native apps to other Snowflake accounts. This is the most complete mapping to the scenario.
- D. Incorrect.
Incorrect. Streams and Tasks are used for change tracking and orchestration/automation, not for building a customer-facing UI. Cortex Search is designed for search and retrieval use cases, not for training forecasting models. Virtual warehouses provide compute and cannot be shared as an application distribution mechanism. This option combines real Snowflake features in incorrect roles.