Prompt Engineering for Java Programmers

Format: 3-Day Intensive Workshop 

Target Audience: Mid-level Java developers 

Primary Goal: Use LLMs as a developer productivity tool and make first steps toward integrating LLM capabilities into Java applications

Prerequisites

Participants should arrive with:

• Solid Java proficiency (comfortable with OOP, collections, exceptions, and standard library)

• Familiarity with REST APIs and JSON

• Basic IDE fluency (IntelliJ or Eclipse)

• Exposure to unit testing (JUnit or TestNG)

• A working Java 17+ environment with Maven or Gradle configured

• API key for Claude or OpenAI provisioned before class (instructor provides setup guide in advance)

No prior AI/ML background required. No Python required.

Learning Objectives

By end of the course, participants will be able to:

1. Explain how LLMs process input and produce output using concepts familiar to Java developers

2. Write effective zero-shot, few-shot, and chain-of-thought prompts for common development tasks

3. Use LLMs confidently to accelerate code generation, refactoring, test writing, and legacy code comprehension

4. Control LLM output format and validate structured responses for use in Java workflows

5. Call an LLM API from Java and handle responses programmatically

6. Build a simple prompt-driven Spring Boot service with conversation history management

7. Recognize and mitigate the most common prompt failure modes

8. Apply a repeatable prompting workflow within a team environment

Module 1: LLMs for Java Developers

• The shift from deterministic to probabilistic systems

• How LLMs differ from APIs, rule engines, and method calls

• Tokens, context windows, temperature, and key parameters — mapped to Java developer mental models

• Lab: Your first LLM API call from Java

Module 2: Anatomy of a Prompt

• System prompts, user messages, and assistant turns

• Roles, personas, and instruction framing

• Positive vs. negative instructions

• Why clarity and specificity matter — and why Java devs have a natural advantage

• Lab: Dissecting and improving poorly written prompts

Module 3: Core Prompting Techniques

• Zero-shot, few-shot, and chain-of-thought prompting

• Iterative prompt refinement

• Controlling output format (JSON, plain text, structured data)

• Lab: Extract structured JSON from unstructured text using few-shot prompting

Module 4: Prompting for Daily Java Development

• Code generation and refactoring

• Writing and improving unit tests (JUnit 5, Mockito)

• Explaining legacy code and generating Javadoc

• SQL and JPA query generation

• Lab: Refactor a legacy Java class and generate a test suite with LLM assistance

Module 5: Calling LLM APIs from Java

• Anthropic and OpenAI SDK overview

• Making API calls, handling responses, and managing errors

• Parsing and validating structured LLM output in Java

• Lab: Build a reusable Java LLM client wrapper

Module 6: Building a Simple Prompt-Driven Service

• Introduction to LangChain4j or Spring AI (one framework, not both)

• Managing conversation history and multi-turn context

• Basic prompt templating

• Lab: Build a simple prompt-driven Spring Boot endpoint with stateful conversation

Module 7: Prompt Failure Modes and Defensive Prompting

• Hallucination — causes, patterns, and practical mitigation

• Prompt injection — the SQL injection of LLMs

• Output validation strategies

• Lab: Intentionally break a prompt, then harden it

Module 8: Testing and Evaluating Prompts

• Why traditional unit testing doesn't fully apply

• Deterministic assertions where possible

• Introduction to LLM-as-judge evaluation

• Lab: Write a basic prompt evaluation harness in Java

Module 9: Prompt Workflow and Team Practices

• Version-controlling prompts like code

• Prompt reuse and building a team prompt library

• Practical guidance on cost, latency, and when not to use an LLM

• Discussion: integrating LLM tooling into your existing Java development workflow

Capstone

• Participants apply the full course workflow: write, test, harden, and evaluate a prompt pipeline for a realistic Java development scenario

• Group debrief