Develop and Deploy LLM Applications GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   A developer needs to build a simple chain that takes a user query, retrieves relevant documents from a vector store, and then passes the query and documents to an LLM to generate an answer. Which component should be used to connect the retriever and the LLM in a LangChain chain?

   • ARetrievalQA
   • BPromptTemplate
   • CLLMChain
   • DConversationBufferMemory
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   When serving an LLM application that uses a Foundation Model API, what is the primary benefit of using Databricks Model Serving?

   • AIt automatically scales the underlying infrastructure based on request load.
   • BIt provides a built-in vector database for retrieval.
   • CIt eliminates the need for any code in the application.
   • DIt guarantees zero latency for all requests.
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A data science team is building a customer-support chatbot that must first extract the user's intent from a free-text query, then call a language model to generate a response, and finally redact any phone numbers from the response before returning it to the user. The team wants to package this as a single MLflow model that can be served as an endpoint. Which approach should they take?

   • AUse mlflow.models.Model with a custom flavor and define a predict() method that only calls the LLM, leaving preprocessing and postprocessing to the serving infrastructure.
   • BCreate a Python model class that inherits from mlflow.pyfunc.PythonModel, implement the predict() method to call preprocessing, the LLM, and postprocessing, and log it with mlflow.pyfunc.log_model().
   • CDeploy a separate preprocessing function as a serverless function and call it from the client before invoking the LLM endpoint, then postprocess the response on the client side.
   • DLog a simple chain using mlflow.langchain.log_model() with a LangChain pipeline that includes a prompt template and the LLM, but skip the redaction step because it is not part of the model.
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   In a LangChain expression language (LCEL) chain, which operator is used to pass the output of one component as the input to the next component?

   • A|
   • B>>
   • C&
   • D+
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A team is building a text-classification model that must preprocess input by removing stop words and postprocess by mapping the model's output scores to labels. They plan to use a pyfunc model for batch inference on a large dataset. What is the best way to implement this?

   • AImplement preprocessing in a separate UDF and postprocessing in another UDF, then call the model endpoint directly from the notebook.
   • BPreprocess the data in a SQL query, call the model via ai_query(), and postprocess the results in another SQL query.
   • CUse a simple chain with LangChain that includes preprocessing and postprocessing, and log it as a pyfunc model.
   • DImplement the preprocessing and postprocessing logic inside the predict() method of a pyfunc model, then use ai_query() to call the model on the dataset.
   Reveal Solution Discussion Select an answer first