# Examples

- `multihop` - Simple as possible demonstration of a graph with two partitions alpha and beta, where beta depends on alpha.

## Ideas

### E-Commerce Search Ranker
- Search to detail page behavior to conversion to return pipeline
- Add in "profile resolution" step to capture dep type? (e.g. person signs in at checkout)
- "input" partitioning: date, country (different infra/laws/products/culture), customer? (what if its for shopify)
- search ranking modeling flow: TDC, train, inference (simple linear model) 
- cross-customer product fraud alliance to qualify products (excuse to have fan-in-and-out pattern - "opt in")
- sessionizing/preprocessing impressions
- returns happen later, and handling them in an atomic, final way
- performance metrics aggregates
- what does multiple paths back to the same partition?