Intermediate 6 topic areas 63+ exercises

Prompt Engineer

Prompt Engineers design, test, and optimise the instructions that guide large language model behaviour. Their daily English covers writing and documenting prompt templates, explaining chain-of-thought reasoning to stakeholders, communicating output quality trade-offs to product teams, and writing evaluation criteria for prompt changes. This path builds the vocabulary for discussing prompt design patterns, output quality, and the language of LLM-powered system design.

Topics covered

  • Prompt design patterns
  • Chain-of-thought reasoning
  • Few-shot learning
  • System prompt architecture
  • Prompt versioning
  • Output quality evaluation

Vocabulary spotlight

4 terms every Prompt Engineer should know in English:

chain-of-thought n.

A prompting technique that instructs the model to reason step-by-step before giving a final answer — improving accuracy on complex reasoning tasks

"Adding 'think step by step' to the prompt improved the model's accuracy on multi-step maths problems from 43% to 78%."
few-shot prompting n.

Providing a small number of input-output examples in the prompt to demonstrate the desired response format or behaviour without fine-tuning

"Three few-shot examples were enough to get the model to consistently output structured JSON in our required format."
system prompt n.

The initial instructions provided to an LLM before a user conversation begins — defining the model's persona, constraints, output format, and task scope

"The system prompt specifies that the assistant should never discuss competitor products and always respond in the user's language."
temperature n.

A parameter controlling the randomness of LLM outputs — lower values produce more deterministic responses, higher values increase creativity and variation

"We set temperature to 0.1 for the classification task where consistency matters and 0.8 for the creative writing feature."
Open full glossary →

📚 Vocabulary Reference

Key terms organised by category for Prompt Engineers:

Prompting Techniques

zero-shotfew-shotone-shotchain-of-thoughttree of thoughtReActself-consistencymeta-promptingrole promptinginstruction following

Prompt Components

system promptuser promptassistant turnpersonaconstraintoutput formatdelimitercontext injectionexampleschain prompt

Parameters

temperaturetop-ptop-kmax tokensstop sequencefrequency penaltypresence penaltyseedlogprobsstreaming

Quality & Versioning

prompt versioningprompt registryevaluation harnessregression testoutput qualityhallucination raterefusal ratelatency trade-offtoken costprompt injection
Study full vocabulary modules →

Recommended exercises

Real-world scenarios you'll practise

  • Writing a system prompt specification document: explaining persona, constraints, output format requirements, and the rationale for each design decision
  • Presenting prompt A/B test results to a product team: explaining what changed, why the metrics improved, and what edge cases remain
  • Explaining chain-of-thought prompting to a stakeholder who wants faster, cheaper LLM calls and is pushing back on the longer reasoning steps
  • Writing a prompt versioning changelog: documenting what changed between prompt versions and the evaluation results that justified each change

Recommended reading

Explore another role

🧬 Synthetic Data Engineer

Open path →