Recommender Systems
Dynamic Queue Adaptation
A deterministic offline prototype for intent-aware music recommendation.
A Python project testing whether manual queue insertions can improve short-term recommendation intent alignment compared with a static seed-only baseline.
Role
Designed the evaluation flow, implemented the reranking logic, and committed deterministic artifacts for reproducibility.
Stack
Links
Question
Can a recommender adapt to a listener's short-term intent when the listener manually inserts tracks into a queue, instead of relying only on the original seed context?
Approach
The project compares a static baseline against an adaptive reranker using offline evaluation scenarios. It avoids external keys and stores outputs so results can be inspected and reproduced.
Result
The adaptive reranker improved intent alignment in 3 of 4 tested scenarios, making it a useful prototype for exploring queue-aware recommendation behavior.