Design comment-likelihood prediction platform

You are designing an ML platform service that scores each post with the probability that a given user will leave a comment. The predictive model itself can be treated as a black box; do not spend time choosing a model architecture. Focus instead on the surrounding ML infrastructure: feature pipelines, feature store, training and serving, real-time inference, and production reliability. Your design must support: Generating training datasets and training models offline. Online…