Ayan Gupta and Rory present hands-on generative AI techniques for Java developers. Learn LLM completions, chat interactions, RAG, and AI function calling with practical code and clear explanations.

GenAI for Java Developers 2: Core Techniques Explained

Presenters: Ayan Gupta, Rory
Resource Links:

This session is a practical walkthrough of the foundational building blocks in creating generative AI applications for Java. Building on the previous setup session, Rory demonstrates how to use multiple techniques that unlock powerful AI-driven features for your applications. The content covers:

Table of Contents

  1. Introduction: Technique Diversity
  2. Setting Up the Environment
  3. Technique 1: LLM Completions
  4. Technique 2: Multi-Turn Chat
  5. Technique 3: Interactive Chat
  6. Technique 4: Retrieval Augmented Generation (RAG)
  7. Technique 5: Function Calling
  8. Session Recap

Introduction: Technique Diversity

This episode uses the coffee machine metaphor: each “brewing button” on the GenAI platform represents a different method for interacting with LLMs in your Java applications. Just as espresso, latte, and cappuccino result in different drinks, each AI technique produces unique application behavior.

Setting Up the Environment

  • Using GitHub Codespaces for demonstration and code execution.
  • Continuation from session 1 setup: ready-to-run sample code.

Technique 1: LLM Completions

  • Simple one-turn interactions where the model provides a completion.
  • Example: Feeding a prompt and receiving a single response from the model.

Testing Simple Completions

  • Step-by-step guide on how to send prompts and handle model replies.

Technique 2: Multi-Turn Chat

  • Building conversations that maintain context and history across multiple user messages.
  • Java code examples show how to manage chat sessions for continuous interactions.

Understanding Chat History and Context

  • Techniques for ensuring responses remain relevant based on conversation flow.

Technique 3: Interactive Chat

  • Creating chat applications that update in real-time as users interact.
  • Live coding example for building an interface that responds to input dynamically.

Technique 4: RAG (Retrieval Augmented Generation)

  • Implementation of RAG to ground AI responses using external documents.
  • Minimizing model hallucination by referencing specific data sources.
  • Step-by-step instructions to integrate document retrieval with Java AI workflows.

Preventing Hallucinations with RAG

  • Explanation of why RAG reduces hallucination and how to add RAG to AI-powered solutions.

Technique 5: Function Calling

  • Integrating AI with external business logic like weather APIs or calculators.
  • Building, calling, and connecting functions within AI-driven Java apps.
  • Practical demonstration of calling weather and calculator functions from the LLM.

Session Recap

  • Review of each AI technique: why it matters and when to use it.
  • Roadmap for next episode: building real-world applications using all covered techniques.

By the end of this session, developers will be equipped with practical patterns and code samples to experiment with completions, conversational AI, RAG, and integrating AI with traditional application logic in Java.

Next Steps:

  • Explore the provided resource links to access code samples and additional tutorials.
  • Prepare for the next video, where the techniques are combined to build complete generative AI applications.