prompt-engineer

Design LLM prompts for multi-step tasks by applying patterns like few-shot, chain-of-thought, and ReAct, then validating them through repeatable guardrails and measurable evaluation criteria.

Format LLM outputs as structured JSON, XML, or markdown by defining explicit output format specifications within the system prompt to ensure consistent and parseable generation results.

Evaluate prompt quality and consistency by applying measurable evaluation criteria, conducting consistency testing, and performing edge case analysis across various LLM interaction scenarios.

Use chain-of-thought and self-consistency prompt patterns to improve LLM output robustness, reduce token usage, and enhance overall prompt clarity for complex reasoning tasks.

Implement guardrails for LLM prompt deployment when building robust multi-step tasks, agent orchestration, and governance workflows that require repeatable constraints and measurable evaluation criteria.

System prompts fail to enforce role definition and output constraints when they lack clear task definitions, explicit output format specifications, and repeatable guardrails for testing.