Java Meets AI: A Hands-On Guide to Building LLM-Powered Applications with LangChain4j
Conference (INTERMEDIATE level)
Room 6
Are you interested in building applications powered by Large Language Models (LLMs) using Java and Spring Boot?
You can create your own AI-powered chatbots, process loads of unstructured data, and automate processes with the help of autonomous agents that can use various tools: call APIs, access databases, and even dynamically execute generated code!
In this talk, we will cover everything you need to know to build your own LLM-powered app. We'll start by exploring the basic building blocks: various commercial and open-source LLMs (OpenAI, HuggingFace, etc.), document loaders, embeddings, numerous vector stores (Pinecone, Elastic, etc.), memory, agents, and tools. We'll then demonstrate how to easily chain these blocks together using a concise and unified Java API.
We'll put LangChain4j into action by building a highly patient customer support agent that handles bookings, cancellations, and provides answers personalized to the customer and tailored to the company's policies.
To help you get started with your own apps, we'll discuss how to select the right LLMs, embeddings, and vector stores for your specific use case, as well as the trade-offs to be made. We'll also cover how you can improve the quality by adjusting parameters, pre-processing your data, and crafting efficient prompts.
By the end of this talk, you'll be able to build an LLM-powered app using Java, and you'll know how to choose the most suitable components for your specific requirements.
You can create your own AI-powered chatbots, process loads of unstructured data, and automate processes with the help of autonomous agents that can use various tools: call APIs, access databases, and even dynamically execute generated code!
In this talk, we will cover everything you need to know to build your own LLM-powered app. We'll start by exploring the basic building blocks: various commercial and open-source LLMs (OpenAI, HuggingFace, etc.), document loaders, embeddings, numerous vector stores (Pinecone, Elastic, etc.), memory, agents, and tools. We'll then demonstrate how to easily chain these blocks together using a concise and unified Java API.
We'll put LangChain4j into action by building a highly patient customer support agent that handles bookings, cancellations, and provides answers personalized to the customer and tailored to the company's policies.
To help you get started with your own apps, we'll discuss how to select the right LLMs, embeddings, and vector stores for your specific use case, as well as the trade-offs to be made. We'll also cover how you can improve the quality by adjusting parameters, pre-processing your data, and crafting efficient prompts.
By the end of this talk, you'll be able to build an LLM-powered app using Java, and you'll know how to choose the most suitable components for your specific requirements.
Lize Raes
LangChain4j
Lize Raes is a Java software engineer with a background in electrical engineering. She began her career in cochlear implant research at Ghent University, earning a SWIFT award in 2011. During the COVID-19 outbreak, she developed a prognosis model and advised the Belgian government together with her team. Currently, Lize is working in Switzerland, where she develops software for drug discovery and gene technology. In her spare time, she is a core member of the team driving LangChain4j developments.