AI Chatbot Prompt Optimization
Budget: $250 – $750 USD
I need an experienced prompt engineer to help me tighten the way my technology-focused chatbot answers questions so that each reply is factually spot-on. Right now the model occasionally drifts or confuses similar concepts; your job is to craft, test, and refine the prompts so the accuracy of its information is consistently high across hardware, software, and emerging-tech topics.
You’ll have direct access to the current prompt set, sample user queries, and a lightweight evaluation script (Python, OpenAI API). I’m looking for someone who can iterate quickly, run controlled A/B tests, and clearly explain the rationale behind each modification. Familiarity with GPT-4, system/user/assistant role structuring, temperature and top-p tuning, and chain-of-thought techniques will be important.
Deliverables
• A library of optimized prompts (system and user variants) covering core technology categories
• A brief report summarizing test methodology, accuracy metrics, and recommended deployment settings
• Guidance on how I can maintain the same accuracy level as models, knowledge bases, or my product scope evolve
If you’ve successfully boosted factual precision for other tech chatbots, I’d love to see an example.
You’ll have direct access to the current prompt set, sample user queries, and a lightweight evaluation script (Python, OpenAI API). I’m looking for someone who can iterate quickly, run controlled A/B tests, and clearly explain the rationale behind each modification. Familiarity with GPT-4, system/user/assistant role structuring, temperature and top-p tuning, and chain-of-thought techniques will be important.
Deliverables
• A library of optimized prompts (system and user variants) covering core technology categories
• A brief report summarizing test methodology, accuracy metrics, and recommended deployment settings
• Guidance on how I can maintain the same accuracy level as models, knowledge bases, or my product scope evolve
If you’ve successfully boosted factual precision for other tech chatbots, I’d love to see an example.