From problem-solving and prompt design to output evaluation and building a maintainable prompt system.
A practical Prompt Engineer isn't just about writing good code. This role requires understanding the problem, data, evaluation criteria, and how to integrate prompts into a system that can monitor them.
Weeks 1–2: Foundations of communication with the model
Practice describing objectives, providing examples, and specifying formatting requirements. Try the same task on multiple prompt variations and note the differences instead of relying on intuition.
Weeks 3–4: Structured output and tool use
Learn JSON schema, validation, and how to handle incorrect model output. Build a small flow that allows the AI to select a tool to read data but not yet perform a write action.
Weeks 5–6: Retrieval and Context
Practice splitting documents, finding relevant sections, and requesting answers with sources. Practice distinguishing between errors caused by retrieval and errors caused by inference models.
Weeks 7–8: Evals and Portfolio
Create a set of 30–50 test cases, scoring criteria, and baselines. Complete a project with a README describing the problem, architecture, risks, measurement results, and design decisions.
