Sustainability-focused Agentic RAG Assistant
The project focused on automating the extraction of GHG emissions disclosures from financial institutions’ annual reports, replacing manual processes. A retrieval-augmented generation (RAG) system was developed and tested for this purpose.
The students worked on this project for ING, competing as a team to develop the best solution, alongside an ING team who were also tackling this project.
The students successfully streamlined the extraction of GHG disclosures and implemented both text and table extraction strategies. A highly accurate retrieval system and robust generation process were developed, producing structured outputs suitable for downstream applications such as audit and analytics. Additionally, an advanced multi-agentic setup was implemented for the RAG architecture.