Jerry Liu, co-founder of LlamaIndex, teaches how to build research-assistant agents that reason over documents in multiple steps, compose tools like summarizers and vector search, and infer metadata filters for precise retrieval.
I'm thrilled to introduce Building a gentic rag with Llama Index. Built in partnership with llamaindex and taught by his CEO Jerry Liu, this course walks you through how to build a research assistant agent that can reason over your documents and answer complex questions. The standard rag pipeline lets you load index and query over your data in a single turn or kind of in a single prompt and response. This works for simple questions, but Jerry will show you how to build agentic ingredients that can do much more complex multi step reasoning. This includes routing to dynamically select from different tools like a summary tool that summarizes documents as an intermediate step to further processing or vector search to help retrieve more relevant context. You then also learn how to generalize this to tool use, creating an interface for agents to to call a rich family of tools and also specify the right arguments. We can apply this to have LLMs figure out how to call a vector database. For instance, if a query requires information from a specific paragraph or page in your document, the LLM can infer the right metadata filters to get back the results you need. I'm delighted to introduce the instructor Jerry, who's co founder and CEO of Llama Index.
Thanks, Andrew. In this course, you'll start with the basics of agentic rag and jump immediately to building your own research agent that can reason over multiple documents. For example, you'll see how to build an agent that can develop and execute a plan to retrieve information from multiple archive research papers, summarize them and even compare them. Controlling and debugging agents is also an important topic. With highly autonomous agents that decide by themselves what actions to take and what tools to use, how do you steer them? You'll also learn best practices for controlling and debugging agents in a granular way.
One of the most rapidly growing applications of agentic AI right now is building research agents. I think you find these techniques powerful and exciting.
Please sign up and let's get started.
