LuLarge – Trustworthy ESG Data Extraction for GSSS Bonds
Co-developed a production-ready, end-to-end prototype that extracts trustworthy, structured ESG data from unstructured sustainability reports for the GSSS bond market.
Key contribution
- Co-built the retrieval pipeline ("Eye") and the extraction/summarisation engine ("Brain")
- Built transparency and provenance visualisation to trace extracted metrics back to their source passages
- Established manual ground-truth datasets and a continuous benchmarking setup for QA
Outcome
A scalable workflow for traceable ESG data extraction, improving comparability and transparency for compliance-oriented workflows.
Focus
- Data Engineering
- ESG / GSSS Bonds
- RAG & LLM
- Provenance & Auditability
- QA & Benchmarking