You will optimize the retrieval and knowledge layer for a vertical: how a domain's content is indexed, linked into entities, ranked, and continuously refreshed, then measured and improved against rigorous IR metrics. This is an information-retrieval and systems role spanning indexing internals, hybrid retrieval, and entity resolution, with the goal of making each vertical the best place in the world to search its domain.
In this position, your responsibility will be to
Design, implement, and operate the retrieval system for a search vertical
Connect and tune the data pipeline, from ingestion to relevance tuning
Build knowledge-graph and entity-resolution layers: entity linking / NER, ontologies, and graph databases (Neo4j or similar)
Develop structured-extraction pipelines over messy, unstructured domain data
Reason about freshness and trust: model how confident we are in a fact and how stale it has become before we serve it
Define evaluation and quality metrics for relevance and drive measurable improvements
Collaborate with crawling, indexing, and ML teams to ensure retrieval and ranking requirements are met
Enable safe experimentation with retrieval, ranking, and extraction strategies
You may be a good fit if you:
6+ years of software engineering experience, some of it in search / information retrieval
Strong IR fundamentals: inverted indexes, BM25/TF-IDF, query understanding, ranking, and evaluation (nDCG/MRR/recall@k)
Experience with vector & hybrid retrieval: ANN, dense+sparse fusion, embeddings models
Experience building structured extraction over messy/unstructured domain data
Fluent in Python and comfortable with systems-level performance work
