Conversational AI Engineer

I'm looking for someone to fix our AI matchmaking assistant. Members talk to the AI to build a profile and get matched. The problem is the system prompt, its about 8,000 words, ~12,500 tokens, and roughly 102 separate instructions, all loaded on every message. The AI forgets rules, re-asks things already answered, skips her own logic, its a bit of a mess at the moment. I've already paid several people to fix this; shortening or lightly rewriting the prompt didn't work. Some important context: not everything is AI-driven. Matching, rate limits, and a few profile actions run on hard-coded rules in the backend (e.g. matching only starts when someone types an exact phrase and confirms with another exact phrase). That's deliberate , we can't let the AI decide when to match or trigger sensitive actions. Your job is to fix the prompt around that, not replace it with "let the AI decide." What I need: someone to split this prompt into stages, always-on basics (voice, safety, formatting) plus smaller blocks that only load when relevant (building profile, updating location, post-match feedback, etc.). Same bot, same conversation; just stop making the AI read 8,000 words when it only needs 1,500 for that moment. I'm not looking for generic prompt writers or people who've only played with ChatGPT. I need someone with serious experience refactoring large production chatbot/LLM prompts and who understands why monolithic prompts fail. If that's not you, please don't apply. I've already spent too much on the wrong fit. Reply with one similar project you've done and how you'd approach this in plain terms. AI responses to this job post will automatically be ignored. Don't even try it, i will notice instantly.