Agnes van Belle

Realized a neural embeddings technique for semantic query expansion and recommendation in multiple languages. Improved the internal A/B testing framework with a minimum sample size calculation, linearization of nonlinear estimators, etc.

Team lead for the Search R&D team. Improving matching and ranking quality using deep learning and learning to rank techniques. Introduced tracking and feedback collection methods for that. Supervised improvements on document extraction, and consulted other teams on integration efforts.

Improved the retrieval performance and functionality of the search product, including back-end development in Java and C++, with a focus on enrichment and ranking techniques, e.g. I implemented a re-ranking module and wrote Elasticsearch plugins.

Evaluated and used information extraction (entity recognition and linking, relationship extraction) techniques to detect anomalies for more efficient forensic spam and malware investigation.

Realized various classification and document understanding projects, e.g. job category classification for an employment agency; text segmentation on reports of municipality workers; and topic discovery in travel agency reviews.

Student Assistant for programming courses in Java and C++