|
|
Designed and implemented a collaborative-filtering–based user embedding framework with streaming updates to mitigate cold start, enabling effective use in large-scale recommendation and advertising models.
2021-2022Engineered scalable user profiling pipeline that aggregated content labels to the user level—incorporating NSFW filtering, label grouping, and temporal decay—delivered via batch (Airflow/BigQuery) and streaming (Flink) systems, with downstream user-to-subreddit mappings powered by approximate nearest neighbors.
2021-2022