Improve LLM Prompts and Structured JSON Responses

We need an AI Engineer to improve prompts for an existing LLM based feature. The current system uses OpenAI API, but the responses are not always consistent. We need someone to improve the prompt instructions, add structured JSON output, reduce irrelevant responses, and make the results easier to use inside our application. Required experience: • Prompt Engineering • OpenAI API • Large Language Models • Structured JSON Output • Function Calling • LLM Response Accuracy • Context Optimization • Python or Node.js Please share your experience improving LLM prompts, structured outputs, or AI response quality.