• Built a custom NLP named-entity-recognition web scraper, doubling data acquisition throughput via automated collection.
  • Curated a schema.org-derived training dataset, raising NER model accuracy from 50% to ~90% on internal benchmarks.
  • Researched and proposed a graph-database storage stack, reducing data retrieval time by 20% and enabling new structured-knowledge workflows.