Unite! Researcher Discovery
Semantic search and visualization platform for 200k+ researcher profiles across 9 European universities.
A production system that powers researcher discovery across the UNITE! alliance, a network of 9 partner universities spanning Poland, Portugal, Italy, Germany, Spain, Finland, Austria, France, and Sweden. The platform indexes and searches 200k+ researcher profiles with both semantic and keyword matching.
What it is
The system combines vector search (pgvector with HNSW indexing) and BM25-based keyword matching into a hybrid search interface, powered by multilingual embeddings from BAAI/bge-m3. Users can search by research keywords, institution, or via conversational queries (“Find renewable energy experts in Spain with 10+ publications”). The backend streams RAG responses via LangChain integration with GPT-4 and Claude. Interactive UMAP visualizations reveal research clusters and collaboration networks across universities, with export capabilities for grant proposals.
Highlights
- Hybrid search engine: 75% semantic + 25% keyword (configurable) with <200ms latency
- Conversational AI: Natural language queries with multi-turn context and streaming responses (<300ms first token)
- Multi-source data pipeline: Automated ingestion from Scopus and Semantic Scholar APIs with 50K+ indexed publications
- Research visualization: Interactive network plots of 200k+ profiles, color-coded by university
- Production scale: 99.9% uptime, 100+ concurrent users, HNSW indexes for fast vector similarity