NYU Libraries
Built semantic search for NYU's library archives
- Partnered with NYU Libraries' Knowledge Access team to scope and deliver a proof of concept for semantic search over 10K+ scholarly records, leading a 5-person team across a 15-week engagement.
- Ran discovery with 10+ faculty and stakeholders and a competitive analysis of 8+ academic search platforms, turning findings into requirements, relevance criteria, and a 4-stage phased build plan.
- Owned the sponsor SOW, milestone tracker, and risk log, delivering every phase on schedule with recurring status updates to department sponsors.
- Built an end-to-end pipeline that processes scanned PDFs through OCR, with two-stage retrieval using open-source language models, serving results in 250–550ms.
- Designed the evals, defining success criteria and a test query set, then prioritized fixes from failure cases and iterated until 95% of queries returned relevant top-5 results.