🤖 AI & Prompt Engineering

Prompt engineering is the practice of writing precise instructions that guide an LLM to produce a useful, accurate output. For IT professionals, it is a core skill because AI tools are now embedded in code editors, code review pipelines, documentation generation, and support workflows. A well-structured prompt reduces hallucinations and produces outputs that are closer to production-ready.

Zero-shot prompting asks the model to complete a task with no examples provided — you rely entirely on the model's pre-trained knowledge. Few-shot prompting includes one or more worked examples in the prompt before stating the task, which steers the model toward the desired format and reasoning style. Few-shot is more reliable for domain-specific or structured outputs; zero-shot is faster and sufficient for well-understood tasks.

A system prompt sets the model's role, constraints, output format, and tone. A good professional system prompt specifies who the model is ("You are a senior software engineer reviewing a pull request"), what it should and should not do ("Flag bugs and suggest improvements; do not rewrite code unless asked"), and the expected output format ("Respond in bullet points with severity labels"). Clear constraints reduce irrelevant or over-verbose responses.

Hallucination refers to an LLM generating content that is factually incorrect, made-up, or unsupported by the input it was given, presented with apparent confidence. In professional IT contexts this is a critical risk — a model might cite a non-existent library, invent an API endpoint, or produce plausible-sounding but wrong configuration values. Mitigation techniques include grounding responses with retrieved documents (RAG) and asking the model to cite sources.

Chain-of-thought prompting asks the model to reason step by step before giving a final answer, improving accuracy on complex tasks. You can trigger it explicitly with phrases like "Think through this step by step" or "First identify the issue, then propose a solution." This technique is particularly effective for debugging prompts, code review instructions, and multi-step analysis tasks where intermediate reasoning matters.

RAG stands for Retrieval-Augmented Generation — a pattern where relevant documents are retrieved from a knowledge base and injected into the prompt context before the model generates a response. This grounds the model's answers in up-to-date or domain-specific information and reduces hallucination. When writing prompts for RAG systems, you must design the prompt to instruct the model to use only the provided context and signal when it does not have enough information.

Temperature is a parameter that controls the randomness of the model's token selection. A low temperature (closer to 0) makes outputs more deterministic and consistent, which is preferable for code generation, structured data extraction, and factual Q&A. A high temperature (closer to 1 or above) increases creativity and variety, useful for brainstorming or generating diverse suggestions. For most professional IT use cases, a low-to-medium temperature is appropriate.

The context window is the maximum amount of text (measured in tokens) that a model can process in a single request, including both the prompt and the generated response. When designing prompts for large codebases or long documents, you must manage what fits within this limit. Techniques include summarising context, chunking documents, using RAG to retrieve only relevant sections, and prioritising the most important content at the beginning of the prompt.

Prompting instructs a pre-trained model at inference time using natural language; no model weights are modified. Fine-tuning updates the model's weights on a domain-specific dataset, making the model inherently more accurate on that domain without needing lengthy prompts. Prompting is cheaper, faster, and reversible; fine-tuning requires labelled data, compute budget, and ongoing maintenance. For most IT teams, advanced prompting and RAG are the practical starting point before considering fine-tuning.

The AI and Prompt Engineering section on CoderLingo includes five exercise sets: writing clear AI prompts, evaluating LLM outputs using professional vocabulary, core AI and ML vocabulary (LLM, RAG, embeddings, tokens), discussing AI risks and ethics in professional English, and describing AI system design including RAG architectures and vector databases. Levels range from Beginner to Advanced.