Getting Started with LLMs: From Zero to Your First Application
Welcome to this introductory path. Here, you will learn the theory and practice behind Large Language Models (LLMs) step-by-step, so you can build your own AI applications with confidence.
Module 1: Basic Concepts
Discover exactly what an LLM is and understand the fundamentals behind the technology.
Learning Objectives
- Understand how text is split and processed as tokens.
- Explain the mechanics of next-token prediction in your own words.
- Recognize the concept of a context window and the memory limits of models.
Exercise
Use an online token calculator (for example, OpenAI's tokenizer) and enter a complex Dutch sentence. See how the sentence is broken down. Write down how many words you used versus the number of generated tokens, and think about why Dutch words sometimes break down into more tokens than English words.
Module 2: Prompting
The way you ask a question (the prompt) directly determines the quality and relevance of the response.
Learning Objectives
- Apply the difference between zero-shot (no examples) and few-shot (a few examples) prompting.
- Assign specific roles and constraints using a powerful system prompt.
- Break down complex tasks into clear, structured instructions.
Exercise
Write a few-shot prompt to train the AI to extract the sentiment (positive, neutral, or negative) from three fictional customer reviews, including a JSON output structure as an example. Test your prompt in a freely available chat interface.
Module 3: API Basics
Step out of the web interface and learn how to communicate with LLMs programmatically via code.
Learning Objectives
- Understand what an API endpoint is and how authentication works.
- Safely store and manage API keys (never in public code).
- Structure a standard JSON payload (including a messages array) for an LLM request.
Exercise
Write (or generate using AI) a simple cURL command or Python script that sends a basic "Hello world" query to an LLM API endpoint (such as Anthropic or DeepSeek) and study how the returned JSON response is structured.
Module 4: Safety & Privacy
Deploying AI responsibly is the most important step before going live. Prevent data leaks and unreliable output.
Learning Objectives
- Filter Personally Identifiable Information (PII) from user input.
- Understand the causes and risks of hallucinations (making up facts).
- Understand the high-level concepts of Retrieval-Augmented Generation (RAG) as a method to enforce factual answers.
Exercise
Write a fictional customer complaint email, including name, address, and account number. Then, devise a logic (or a prompt for a pre-processing AI) to completely anonymize this text before it can be sent to an external API endpoint.