Prompt Engineer

Job Description

We’re looking for a Prompt Engineer who combines linguistic precision with computational intuition to craft, refine, and evaluate complex prompt chains for large language models (LLMs). This isn’t a generic “play with ChatGPT” role—our work involves systematizing prompt behaviors for real-world applications, building reusable prompt templates, interpreting model limitations, and collaborating with product teams to ensure the right outputs are delivered with consistency and accuracy. You’ll be expected to think like a developer, write like a copywriter, and test like a scientist.

Key Responsibilities

• Design, test, and optimize prompt frameworks that produce reliable outputs across edge cases and evolving model updates
• Build structured prompt chains that integrate with APIs, retrieval-augmented generation (RAG), and few-shot examples for dynamic use cases
• Collaborate with researchers and data scientists to define model evaluation benchmarks tied to prompt performance
• Interpret token-level output behaviors and apply temperature/top-p sampling knowledge to manage randomness vs. determinism in generation
• Prototype lightweight tooling for prompt iteration, version control, and A/B testing
• Identify prompt-based vulnerabilities such as jailbreaks, prompt injections, or hallucinations, and create mitigation strategies
• Document prompt patterns, model quirks, and failure modes in an internal prompt playbook

Minimum Qualifications

• Strong grasp of LLMs (GPT-4+, Claude, Mistral, Gemini, etc.), including tokenization behavior and context window limits
• 2+ years of experience in NLP, applied ML, or product-focused LLM deployments
• Proficiency with at least one scripting language (Python preferred), especially for working with APIs and model outputs
• Demonstrated experience designing effective zero-shot, few-shot, and instruction-tuned prompts across varied domains
• Excellent writing and editing skills, especially under tight character constraints and structured formatting
• Familiarity with vector databases and RAG pipelines is a plus
• Ability to interpret model misbehaviors and think creatively around prompt debugging
• Experience collaborating across technical and non-technical teams in agile environments

Preferred Qualifications

• Background in linguistics, cognitive science, technical writing, or UX copy
• Experience with prompt engineering tools (LangChain, PromptLayer, Guidance, etc.)
• Published work or thought leadership in LLM prompt design or evaluation

What Success Looks Like

• You’re maintaining a library of tested, high-precision prompt chains reused across products
• You’ve identified model failure points early and tuned prompts to mitigate risk
• You’re driving documentation standards around prompt behavior and tuning principles
• Your work increases model utility, reliability, and safety in measurable ways