Ten working days, end-to-end, from "we want AI in our product" to a working production feature you can put in front of users.

Typical scope: a RAG (retrieval-augmented generation) flow over your documents, an agent workflow with safe tool use, a function-calling integration, or a structured-output pipeline. Stack: Python + LangChain + your LLM provider of choice (OpenAI, Anthropic, open models). Deployment on your existing cloud — Azure, AWS, or self-hosted.

What you get: a working feature integrated into your stack, written architecture notes, a security review (auth, input validation, prompt-injection mitigation, audit trails), and a hand-off doc your engineers can extend from.

Best fit: companies with an AI hypothesis but no in-house LLM experience yet. I take you from "let's try this" to "we shipped this".

Background: AI delivery is the current focus of my freelance practice, plus internal AI enablement and ML pilot projects from my last in-house role.