Transforming Data Interaction: A Self-Served Insight Tool with LangGraph

In today's data-driven landscape, extracting meaningful insights from complex datasets is essential. This talk introduces an innovative project at Merck KGaA's Life Science KGaA that leverages LangGraph—a library for building stateful, multi-actor applications with Large Language Models (LLMs)—to create a self-served conversational agent. This agent enables interactive querying, automatic code generation, and key insights summarization, freeing users from manual analysis and enhancing decision-making.

We will delve into the architecture of the LangGraph agentic framework, showcasing how we integrate various tools to process queries, trigger analytical functions, and present data in user-friendly visual formats. We’ll also share our evaluation experiences using LangFuse and Ragas, emphasizing metrics that ensure optimal performance and user satisfaction for LLM agent projects.

Through real-world applications, we aim to demonstrate how conversational agents can transform data interaction, making customized insights more accessible and intuitive. Participants will gain insights into the potential of LangGraph and practical strategies for evaluating AI-driven data analysis projects, setting the stage for future advancements in this field.