Modular building blocks that turn live behavior, catalog intelligence, and real-time context into adaptive recommendations, search results, and ranking decisions across the user journey.
All primitives are built on the same perception and ranking infrastructure. Customers can use Albatross to generate recommendations, personalize search, or rerank candidate lists from existing systems.
| Primitive | What it does | Primary benefit | Use cases | Input |
|---|---|---|---|---|
Discovery Feed | Adaptive product discovery for every session | Drives discovery, engagement and conversion | Homepage, category pages, purchase-optimized feeds, marketing campaigns | user_id, k |
Similar Items | High-coverage alternatives across large, changing catalogs. | Improves product exploration and substitution | Product detail pages, item alternatives, second-hand marketplace substitution | item_id, k |
Playlist | Sequential recommendations from a set of selected items. | Builds contextual discovery experiences for baskets, collections,and saved lists | Basket building, checkout page, saved items, style boards | list of item_ids, k |
Contextual Search | Search that is aware of the user session and context, not only the query | Turns search into a personalized, adaptive experience | Personalized search reranking, query refinement, multimodal search | query, user_id, k |
Custom Model Components can be added to every primitive.
Examples include optimization for a different business objective (e.g. purchases), or balancing promotions and content types.
These custom solutions are not part of the primitives subscription and are billed separately.
A walkthrough of Discovery Feed, Search, and Playlist running live on a marketplace.