Frontend -> API -> messageQ -> Recommendation Engine (Candidate Generation -> Ranking -> return to another messageQ?).
Everything from creating the messageQ onwards needs to be orchestrated together, and handling the return will be quite interesting. This will all need to be deployed on EC2 as well. API calls to the Rec Engine are made PER SWIPE PER USER.
Requirements
Asynchronous Recommendation generation
Blazing fast probably
Deployment on EC2 probably - different use case than our Data API EC2
Orchestrator and Deployment (12+ Hours)
Background
Frontend -> API -> messageQ -> Recommendation Engine (Candidate Generation -> Ranking -> return to another messageQ?).
Everything from creating the messageQ onwards needs to be orchestrated together, and handling the return will be quite interesting. This will all need to be deployed on EC2 as well. API calls to the Rec Engine are made PER SWIPE PER USER.
Requirements