Databricks Generative AI Engineer Associate Question 88
Select 3You are building a generative AI application on Databricks that uses a fine-tuned large language model (LLM) to provide customer support responses. You want to ensure the application can handle multiple asynchronous user queries efficiently while maintaining session-specific context. Which of the following steps should you implement in your application?
- A
Use a vector database to store session-specific embeddings for context retrieval.
- B
Implement a synchronous REST API to handle incoming user queries.
- C
Leverage Databricks' MLflow to track the fine-tuning process and manage model versioning.
- D
Use distributed Spark jobs to pre-generate responses for all possible user inputs.
- E
Maintain a session-based context mechanism to feed relevant history into the model for each query.
Show answer and explanation
Correct answers: A, C, E
Explanation
To build an efficient generative AI application, you need to ensure session-specific context is maintained for relevant responses (using a vector database and session-based context mechanism). Additionally, leveraging MLflow for fine-tuning and versioning ensures reliability and reproducibility in your application. Avoiding synchronous APIs improves scalability, and pre-generating responses is impractical for dynamic user queries.
- A. Correct.
Using a vector database to store session-specific embeddings is crucial for retrieving relevant context efficiently when handling multiple asynchronous queries.
- B. Incorrect.
A synchronous REST API would not be suitable for handling multiple asynchronous queries, as it could lead to bottlenecks and reduced scalability.
- C. Correct.
MLflow is a best practice for managing the fine-tuning process and model versioning, ensuring the correct model is deployed in the application.
- D. Incorrect.
Pre-generating responses for all possible user inputs is not feasible due to the vast number of potential queries in a generative AI application.
- E. Correct.
Maintaining a session-based context mechanism is essential to ensure the model generates responses that are relevant to the ongoing conversation.