Prompt Engineer Jobs

Prompt Engineer Jobs

Companies across tech, finance, healthcare, and media are hiring Prompt Engineers to design, test, and optimize the instructions that make LLMs and generative AI tools reliable, safe, and useful. If you love language, systems thinking, and rapid experimentation, this role sits at the intersection of product, research, and engineering.

Job Overview

Prompt Engineers build and refine the prompts, system messages, few-shot examples, and evaluation pipelines that guide large language models and multimodal AI. You’ll turn messy product goals into consistent model behavior, reduce hallucinations, and ship features like chatbots, copilots, agents, and content generators. Expect heavy iteration, data analysis, and close work with ML researchers, product managers, and designers.

About the Company

We’re a product-led AI company building tools that help millions create, learn, and automate work. Our stack includes open and proprietary LLMs, RAG systems, and multimodal models. The culture is fast, experimental, and documentation-driven. Engineers here ship to prod weekly and measure success by user impact, not just model metrics.

Prompt Engineer Responsibilities

• Design & iterate prompts for chat, tool use, classification, extraction, and generation tasks across text, image, and code models
• Build eval sets and run A/B tests to measure quality, latency, cost, and safety regressions
• Write guidelines for tone, refusal behavior, citation, and tool-calling logic; partner with policy on guardrails
• Optimize context with retrieval, summarization, and prompt compression to hit quality + cost targets
• Debug failures using logs, traces, and red-teaming; identify edge cases and failure modes
• Collaborate cross-functionally with PM, research, data, and legal to ship features from prototype to GA
• Automate workflows with Python scripts, eval harnesses, and prompt versioning in git

Required Skills

• Systems thinking: Break down vague product requests into deterministic, testable behaviors
• Writing: Precise, unambiguous technical writing. You can explain the same idea to a model, a PM, and a lawyer
• Data analysis: SQL or Python pandas to analyze logs, cluster failures, and track win rates
• Experimentation: Hypothesize, test, measure. Familiar with offline evals and online A/B frameworks
• Tools: Git, Jupyter, LangChain/LlamaIndex or similar, vector DB basics, JSON/Regex
• Collaboration: Give/receive feedback on prompts like code reviews; document decisions

Generative AI & LLM Expertise

• Model behavior: Understand context windows, tokenization, temperature, top-p, stop tokens, and function calling
• Prompting techniques: Zero-shot, few-shot, chain-of-thought, ReAct, tree-of-thoughts, constitutional, and RAG patterns
• Safety: Red-teaming, prompt injection, jailbreak defenses, refusal design, and bias evaluation
• Multimodal: Experience prompting for image generation, code execution, or tool use a plus
• Evaluation: Build golden sets, use LLM-as-judge, and track metrics like helpfulness, groundedness, and toxicity
• Cost/latency: Know when to use a small model + good prompt vs a large model; caching and batching strategies

Qualifications & Experience

• Education: Bachelor’s in CS, Linguistics, Philosophy, CogSci, Math, or equivalent experience.
• Experience: 2+ years in software, ML, data, technical writing, or education. Prior prompt engineering or eval work strongly preferred.
• Portfolio: Examples of prompts you’ve shipped, evals you’ve designed, or products you’ve improved with LLMs. GitHub, blog posts, or case studies work.
• Nice to have: Fine-tuning, RLHF data work, dataset creation, or experience with agents and long-context models.

Salary & Benefits

Benefits: Full medical/dental/vision, 20 days PTO + holidays, 16 weeks parental leave, $2k learning stipend, $500/mo model/API credits, home office setup, 401k match.

Why Join This Team?

Impact: Your prompts go live to millions of users. A 1% quality win moves real metrics.
Speed: We run 50+ experiments per week. You’ll see results, not slide decks.
Growth: Learn directly from research scientists and ship with staff engineers.
Ownership: You own a model behavior area end-to-end, from idea to monitoring.
Mission: Build AI that’s helpful, honest, and accessible, not just impressive demos.

Application Process

1. Apply with resume + 2 prompt samples: one for a chat task, one for extraction/classification. Show before/after and why you changed it.
2. Recruiter screen 20 min: Background, logistics, and comp expectations.
3. Take-home 3 hours: Improve a failing prompt, write an eval, explain tradeoffs.
4. Virtual onsite 3×45 min: Prompt review, data analysis, and cross-functional collab.
5. Offer: We move fast. Target 1 week from onsite to decision.

Related AI Jobs

• ML Engineer, Generative AI: Focus on training/fine-tuning, infra, and serving
• AI Product Manager: Define LLM features, eval criteria, and launch plans
• Conversation Designer: UX + scripting for voice/chat agents
• AI Red Teamer: Adversarial testing and safety policy enforcement
• Data Scientist, LLM Eval: Build metrics, dashboards, and offline eval pipelines

Similar Generative AI Careers

• LLM Agent Developer: Chain tools, memory, and planning for autonomous tasks
• AI Content Strategist: Design system prompts and style guides for brand-safe gen
• Synthetic Data Engineer: Generate and curate training data using LLMs
• Model Behavior Analyst: Diagnose why models fail and propose fixes
• Technical Writer for AI: Document APIs, prompting guides, and best practices

Frequently Asked Questions

• Do I need to code?
  Yes, basic Python for evals and scripting. You won’t build production services, but you should read logs and write notebooks.
• Is this just writing prompts all day?
  No. 50% is experimentation and measurement, 30% is cross-functional alignment, 20% is writing/iterating prompts.
• Will I work on model training?
  You’ll partner with researchers on data and evals, but core training is a separate ML team. Prompt work directly influences what data we collect for SFT/RLHF.
• Remote friendly?
  Yes. US/Canada remote with quarterly onsites. SF, NYC, and Seattle hubs available.
• How is performance measured?
  Shipped improvements to quality, safety, or cost. Defined by A/B tests, eval win rates, and reduction in user-reported issues.