Certified Machine Learning ProfessionalDemo

Question

A Data Scientist is developing a product-recommendation model for an online merchant that proposes additional items from a customer’s current shopping basket. To prevent recommending items already in the basket, the model must access the real-time basket contents during inference. The model will be deployed with Databricks Model Serving. The data scientist now needs to implement real-time feature engineering so that the current basket contents are incorporated into the model’s prediction.

Which approach accomplishes this?

A Create a custom pyfunc model that processes the basket contents as part of the model’s predict() method.
B Write basket data to a feature table and publish to the online store for low-latency access.
C Store basket data in a Delta table and configure automatic feature lookup during model serving.
D Implement feature preprocessing in a separate microservice that feeds the model serving endpoint.

Explanation

A custom MLflow pyfunc model can receive the current basket as request input and run basket-specific preprocessing and filtering within predict() for each inference request. Databricks Model Serving supports custom pyfunc code for preprocessing and per-request logic, making it suitable for dynamic features that exist only at scoring time.

Learn more

Deploy Python code with Model Serving

Certified Machine Learning ProfessionalDemo

AstroTutor

Notes

Readiness

QuestionQ29

Community Discussion