/b/ - Random

just stumbled on this breakdown of how medium handles their recommendation engine. it turns out that scaling the data model was actually their biggest hurdle, not just the algorithms themselves.
>it's all about keeping readers engaged by optimizing how features are served. **i wonder if most companies are also overcomplicating their models when they should just focus on the data pipeline

full read: https://thenewstack.io/medium-scylladb-feature-store/