Sorbobot
Overview
Sorbobot helps researchers, managers, and partners discover expertise within Sorbonne University. Given a research topic or question, the system surfaces relevant researchers, labs, and publications across the university’s broad disciplinary landscape.
Partners
| Institution | Role |
|---|---|
| SUMMIT | Engineering team, infrastructure, data owner |
| SCAI | ML/NLP expertise, LLM pipeline, internship supervision |
My Role
Contributed NLP and LLM architecture expertise (RAG pipeline design), mentored junior engineers, supervised one Master’s intern, and coordinated with a freelance senior data engineer.
Technical Approach
- Data ingestion: researcher profiles, lab pages, publication databases (HAL)
- Indexing: embedding-based vector index over researcher and publication content
- RAG pipeline: user queries matched against the index; retrieved context passed to an LLM for synthesis
- Entity linking: connecting researcher, lab, and topic mentions across heterogeneous sources
Stack: Python · LangChain · LLM APIs · PostgreSQL
Status
🟢 SCAI phase completed — prototype built and validated. Project continues under SUMMIT’s ownership.