•Designed and implemented a brand new LLM-based people disambiguation and lookup product. Worked cross-functionally with Infra, ML, and Annotation teams.
•Built a LLM-based answer summarization module for the Answers product.
– Reproduced, evaluated, and deployed Transformer based slot filling model. Owning the full lifecycle of the model.
– Designed, implemented, and deployed a new Multi-annotator Search Relevance annotation interface. Designing and implementing both the front end (REACT) and backend (Django, Postgres). Worked closely with the annotation team to develop the interface. Surfaced valuable agreement statistics and greatly improved data quality. Implemented Dashboards to visualize data quality and inter-annotator agreement.
– Designed a new Entity Linking annotation interface to collect EL data. Worked closely with in-house annotators, UI/UX team, and potential downstream consumers of such signal.
– Re-designed the internal Knowledge Graph representation and developed a new ontology with annotators and downstream consumers.
– Designed and implemented multiple Entity management UI features.
