Engineering & Craft
The craft behind the systems.
Deep technical notes — RAG vs fine-tuning, vector databases, LLMOps, AI security and compliance, model selection, and the architecture, API, and cloud craft that production AI actually runs on.
Common questions.
When should I use RAG instead of fine-tuning?
Use RAG when answers must be grounded in changing, proprietary knowledge and you need citations. Fine-tune when you need a consistent style, format, or behaviour the base model lacks. Many production systems combine both.
How do you keep AI systems secure and compliant?
Data minimisation, tenant isolation, prompt-injection defences, audit logging, and keeping sensitive data out of third-party model context unless contractually permitted — designed in from the architecture stage, not bolted on.
Have a question for the writers?
We answer founder questions in posts.
Send the question; if it's a good one, the post is yours to share. Either way you'll get a senior reply.