A prompt that works in a demo is not the same as one that holds up on the thousandth request. I design prompts for LLM applications, then build the tests that show, with evidence, where they hold and where they break.

What you get

• Prompts and reusable templates for your task, whether that is generation, classification, extraction, summarization, or reasoning

• An evaluation set with rubrics, reference answers, and edge cases, so quality is measured instead of guessed

• A clear error analysis covering hallucinations, reasoning failures, unsupported claims, and instructions the model ignores

• A short written report on what to change in your prompts, context, and workflow, and why

This is the center of my current work as a senior AI/ML engineer, where I design prompts and evaluation rubrics and trace model failures back to their causes.