Summary

Aggregates internal + LinkedIn data; vector + keyword search w/ query expansion & post-ranking to match operators.

Problem Statement

Finding the right operational talent requires searching across internal project history, LinkedIn profiles, and relationship databases. Without unified search, valuable candidates get missed and searches take hours instead of minutes.

Details

Unifies Salesforce + internal project history with scraped LinkedIn profiles into a single talent index. Query expansion (LLM) broadens skill terms; hybrid retrieval (vector + BM25) plus post-ranking (recency, tenure, must-have filters) yields explainable slates.

Exposes “why-this-candidate” signals and tunable weights. Delivered and in active client use.

Images

What We Learned

Hybrid search (vector + keyword) outperforms either approach alone for talent matching. LLM query expansion captures domain terminology variations automatically. Explainable ranking builds trust with recruiters who need to justify candidate selections. Unifying disparate data sources creates value impossible to achieve with siloed systems.