Prompt Engineer — AI Interaction & RAG Design

Steampunk, Inc.; McLean (VA); On-site; USD 115,000 - 140,000; Full time

Job summary

Steampunk, Inc. seeks a Prompt Engineer to design, test, and refine interaction patterns for generative AI systems across client engagements. This role shapes user experiences and ensures LLM outputs are relevant, safe, and aligned to mission needs.

You will collaborate with AI Developers, LLMOps, Evaluation Scientists, and product/design teams to craft prompts, optimize context, improve retrieval, and create intuitive multi-turn flows that reduce cognitive load for users.

Qualifications

• Bachelor's degree in a related discipline or equivalent experience.
• Demonstrated experience designing and testing prompts for LLMs in real applications.

Responsibilities

• Design, test, and refine prompts, templates, and prompt chains for accurate and usable AI outputs.
• Develop context management strategies and retrieval augmentation patterns.
• Collaborate with developers to implement prompt techniques in RAG pipelines and UX layers.
• Conduct experiments (A/B tests, rubrics) to measure prompt performance and identify improvements.
• Define evaluation metrics with AI Evaluation Scientists for quality, relevance, safety, and reduced hallucinations.
• Create user-centric interaction flows with UX/product teams to guide inputs and reduce cognitive load.
• Translate complex content into structured, LLM-friendly formats to improve reasoning and reduce errors.
• Contribute to safe output behaviors via guardrails and context constraints.

Skills

AI/ML experience

Prompt engineering

Python

Analytical thinking

Education

Bachelor's degree

Tools

LangChain

LlamaIndex

Job description

Steampunk, Inc. seeks a Prompt Engineer to design, test, and refine interaction patterns for generative AI systems across client engagements. This role shapes user experiences and ensures LLM outputs are relevant, safe, and aligned to mission needs.

You will collaborate with AI Developers, LLMOps, Evaluation Scientists, and product/design teams to craft prompts, optimize context, improve retrieval, and create intuitive multi-turn flows that reduce cognitive load for users.