Representation Learning and Feature Engineering Questions

Covers both learned and engineered approaches to represent data for prediction and optimization in production systems. Topics include embedding strategies for categorical and entity features, self supervised and contrastive representation learning, multi task objectives, design of geospatial and temporal features for location aware and time sensitive applications, aggregated and hierarchical features, online versus offline feature computation, feature freshness and latency constraints, interaction between learned representations and manual features, feature store design, and trade offs in memory and compute for real time inference.