AI/ML • 2026
GlobaLens
A news-exploration prototype that combines semantic search, generated summaries and a 3D globe over geotagged events.
My Role
Designed and implemented the ingestion, search and globe interface.
Team Size: 1 person
Problem
A stream of headlines is difficult to explore by topic and geography. Keyword matching alone can miss related events described with different words.
Approach
Built a GDELT ingestion pipeline using BigQuery, a Flask API, embeddings and MongoDB vector search. A React globe connects geographic events with search and summaries.
Impact
Demonstrates a data pipeline and a spatial interface for exploring related news. The recorded similarity threshold is a filtering choice; it does not establish search accuracy, user adoption or current feed availability.
Technologies
ReactThree.jsFlaskPythonMongoDBSentence-TransformersBigQueryVertex AI