cxc.ai

We're building AI-powered tools for the trades — HVAC, plumbing, electrical, and more. Our platform connects technicians in the field, customer service managers in the office, and homeowners in real time. We need a prompt engineer who can turn raw, messy, real-world conversations into structured intelligence that actually helps our customers run better businesses.

What you'll work on

• Transcript analysis. Design prompts that analyze real conversations between technicians, CSMs, and homeowners — surfacing what went wrong, what worked, and how to improve.
• Internal document intelligence. Build bots that read company SOPs, training docs, and service guides — then generate summaries, best practices, and recommendations.
• LLM evaluation. Test and benchmark multiple models (GPT, Claude, Gemini, and others) to find the best fit for each use case in our stack.
• Prompt architecture. Write structured, reusable prompts with dynamic context injection, persona framing, and clean output formatting for real business workflows.
• Continuous iteration. Evaluate real outputs against business goals, identify failure modes, and refine prompts based on actual field usage.

What we're looking for

• Hands-on experience with AI models (GPT, Claude, Gemini, Mistral, or similar) — not just using them, but understanding why they behave the way they do
• Ability to design prompts for multiple task types: classification, summarization, QA, recommendations, and structured extraction
• Strong written English — you write prompts for a living, so clarity and precision are non-negotiable
• Comfortable working with messy, unstructured inputs — real field conversations aren't clean
• Self-directed and proactive — you spot gaps before being asked and bring solutions, not just problems

Nice to have

• Background in or exposure to the home services / field service industry (HVAC, plumbing, electrical, pest control, etc.)
• Familiarity with ServiceTitan or similar field service management platforms
• Experience with evals, structured benchmarking, or red-teaming AI outputs
• Ability to write basic scripts (Python or JS) to automate prompt testing