OpenAI API Prompt Optimization
Budget: £250 – £750 GBP
I'm looking for an expert prompt engineer to troubleshoot and improve an OpenAI API prompt used in AutoMind—an AI-powered vehicle diagnostics tool.
The prompt currently works well for basic reports, but breaks down when multiple engine fault codes are present. It’s supposed to group fault codes by component (e.g. turbo, injectors) and generate up to 10 grouped assessments per report. Right now, it often returns only 1 or 2 groups, even when more are expected.
What I need:
- Expert review and refinement of the existing prompt
- Reliable multi-group generation logic (up to 10 fault groups)
- Clear documentation or guidance for future improvements
You’ll be working with:
- Fault code inputs from vehicle ECUs
- A handful of structured user-submitted symptoms
- An AI system designed to explain, assess urgency, and suggest fixes in layman’s terms
Ideal Skills:
- Proven prompt engineering experience with OpenAI (especially GPT-4 or GPT-4o)
- Comfortable working with logic-heavy or multi-step generation prompts
- Experience generating long-form or structured outputs via API
If you've fixed prompt performance issues like this before—or you’re confident you can—please get in touch with your approach and any relevant examples.
The prompt currently works well for basic reports, but breaks down when multiple engine fault codes are present. It’s supposed to group fault codes by component (e.g. turbo, injectors) and generate up to 10 grouped assessments per report. Right now, it often returns only 1 or 2 groups, even when more are expected.
What I need:
- Expert review and refinement of the existing prompt
- Reliable multi-group generation logic (up to 10 fault groups)
- Clear documentation or guidance for future improvements
You’ll be working with:
- Fault code inputs from vehicle ECUs
- A handful of structured user-submitted symptoms
- An AI system designed to explain, assess urgency, and suggest fixes in layman’s terms
Ideal Skills:
- Proven prompt engineering experience with OpenAI (especially GPT-4 or GPT-4o)
- Comfortable working with logic-heavy or multi-step generation prompts
- Experience generating long-form or structured outputs via API
If you've fixed prompt performance issues like this before—or you’re confident you can—please get in touch with your approach and any relevant examples.