When LLM features hit production, the prompts in your codebase become load-bearing infrastructure. They deserve the same rigour as the rest of your system.

• System prompt design: structure, instruction hierarchy, role framing
• Eval-driven prompt iteration: measure before you change
• Prompt versioning and rollout strategies
• Structured output, JSON-mode and tool-use prompt patterns
• Few-shot example curation and test-set design
• Failure-mode analysis: hallucination, prompt injection, edge cases

This isn't prompt-engineering-for-business-users training. It's for engineering teams shipping LLM-powered features who want their prompts treated like code: reviewed, tested and owned.

If your team has prompts living in Notion docs and no one's quite sure when they last changed or how to tell if a tweak made things worse, this is for you.

A few ways we can work together:

• Team workshops. Bring your engineering team up to a shared standard on production prompt engineering.
• Hands-on pairing. Work alongside your engineers on real prompts and evals in your codebase.
• Ongoing review. Recurring sessions to keep your prompt practices sharp as models and tooling evolve.

I bring direct, current experience from shipping production AI, including building Footstep.