Prompting Engineer
Murphy
EUR 35k-45k / year + Equity
Posted on Jul 1, 2025
🧮
Prompting Engineer
Apply now
👋 About us
Murphy AI is a next-generation debt collection platform powered by artificial intelligence, designed to optimize recovery rates while maintaining respectful and personalized communication. Our advanced automation streamlines the process of collecting overdue invoices for businesses, providing a seamless and effective solution.
Our platform integrates effortlessly with ERP systems or supports CSV uploads to ensure smooth access to data on unpaid invoices. Using AI-driven agents, we engage with debtors across multiple channels, including SMS, email, phone calls, and more. For cases where payment arrangements are unsuccessful, we facilitate escalation to legal proceedings through trusted partnerships with law firms.
As a fast-growing startup that has already made an impact within just a few months in the market, we are building a talented team to scale our operations and drive our vision forward. Join us in redefining debt recovery with innovation and empathy.
Type: Full-time, Hybrid in Barcelona
Team: Operations (working closely with AI, Product & Data)
🧠 About the role
As a Prompting Engineer, your mission is to design, evaluate, and iterate the business-facing prompts that power our AI agents’ conversations with users across channels like voice, SMS, and email.
This role goes beyond writing: you’ll own the full prompt evaluation pipeline, including:
The design and maintenance of counter agents to test prompt performance before going live.
The use of LLMs as judges to assess quality, tone, and outcome alignment.
The creation of evaluators to continuously monitor real interactions and flag issues.
You will play a key role in shaping Murphy’s messaging strategy and directly impacting our Success Rate, the core metric that drives our business.
📝 Responsibilities
Prompt Design & Iteration: Write and refine business-layer prompts across all communication channels.
Prompt Analysis: Measure how each prompt variation impacts success rate, engagement, and user behavior and then optimize accordingly.
Prompt Validation Pipeline:
Design and run counter-agent simulations to test prompts pre-launch.
Apply LLM-as-a-judge frameworks to evaluate clarity, empathy, and accuracy.
Build and deploy evaluators to monitor production conversations and detect regressions.
Experimentation: Launch A/B or multivariate tests to evaluate tone, flow, and logic.
Cross-Functional Collaboration:
With AI Engineers to understand prompt behavior at execution.
With Product to ensure alignment with user journeys.
With Data to define and monitor prompt KPIs.
Continuous Improvement: Review real conversations weekly and iterate based on evidence.
📚 Requirements
Experience: 2+ years working with prompts, LLMs, conversational UX, or evaluation of AI outputs and Data Analytics.
Data Orientation: Confident using prompt-level KPIs and conversation data to drive structured improvement.
Initiative: Strong ownership mindset—you fix things before being told.
Writing Skills: Clear, concise, and strategic writing tailored to drive real business outcomes.
Hands-On Python Experience: Able to write clean, efficient code to test prompts, simulate agent behavior, and automate QA workflows.
Business-Driven Thinking: You understand how messaging impacts recovery, conversion, and user trust—and you design with that in mind.
🧩 Nice to have
Familiarity with tools like ElevenLabs, LangFuse, OpenAI, or other LLM frameworks.
Experience with voice-based agents.
Background in UX writing, linguistics, or computational language.
Exposure to regulated industries such as finance, collections or insurance.
🤝 What we offer
💰 Salary: €35,000–€45,000 gross/year
📈 Stock options: You’ll own part of Murphy’s journey, because you're helping build it.
🧠 Impact: Be in the inner circle of a company solving real-world problems with cutting-edge AI.
🚀 Learning: Work across departments and gain exposure to top-tier investors and operators.
🍱 Healthy meals: Enjoy daily meals via Nora Foods, 50% covered by us.
🚴♂️ Panot e-bike: 50% of your subscription covered, so you can ditch the car.