Deep Research Multi agent for a leading investment firm

We built an agentic research system for a leading investment firm in Europe that enriches reports by doing iterative research. It autonomously analyzes documents, reasons around critical questions about potential gaps, and performs structured research to answer them. This uncovers overlooked insights, more recent developments. etc providing a more complete picture for high-stakes decision-making.

Background

Critical business decisions rely on detailed reports, like investment analyses or legal briefs. However, these documents often contain hidden gaps, outdated information, or unstated assumptions. Acting on this incomplete picture creates significant risk and can lead to missed opportunities. Manually enriching these documents is too slow and expensive to be practical.

Solution

We built an autonomous AI agent that reads a document and intelligently generates research questions to find missing information. It then uses various tools tools to investigate these questions across the public web and internal knowledge bases. The final output is a single, consolidated report that merges the original content with the newly discovered insights grounded with references, giving users a more complete and reliable basis for action.

ML tools 

The system uses Large Language Models (LLMs) within a custom multi-agent architecture. It employs the ReAct (Reason + Act) framework, allowing agents to reason and use tools like web search APIs. internal Retrieval-Augmented Generation (RAG) service, and a generation tool for follow-up questions. Built entirely in Python, the system ensures transparency, traceability, and flexible citation management throughout the research process.