Sometimes I get messages asking if I take sponsors. It’s not something I had considered but after some thought, it makes sense. I’m opening up a few sponsorship slots on the blog. A quick bit…
Last updated: 28 July 2026 I’ve been building out an AI agent, Iron Mind AI, a personal trainer agent, incrementally using the Microsoft Agent Framework. Each post in this series adds a new capability: function…
Last updated: 28 July 2026 Retrieval-Augmented Generation is a pattern, not a framework, and the gap between a RAG tutorial and a RAG pipeline that stays healthy in production is substantial. This series covers what…
Microsoft awarded me the Most Valuable Professional Award for the 8th year in a row. The disk was shipped and each year it still means something. I don’t take it for granted. Eight years is…
I’ve been using Claude Code heavily across AI engineering work: RAG pipelines, .NET MCP servers, content automation, agentic workflows. The way I use it now looks nothing like how I started. Most write-ups on Claude…
A continuous Azure WebJob stopped mid-run with no exception, no stack trace, and no shutdown message. Here’s why the Web App memory chart lied, and how to find out who killed your process.
A RAG chunker’s “is this small enough to skip splitting?” check ran against placeholder text, not the real code it stood in for — so oversized code-heavy chunks sailed straight past the embedding model’s token limit.
A production RAG agent threw intermittent 500s because nobody guarded the upper length of the embedding query – here is the two-tier fix and the multi-turn edge case that hides it.
When you build and ship a RAG pipeline, you quickly discover that the real work is not the initial vectorisation pass. It is keeping the pipeline healthy after that. Documents get deleted, re-crawled content fails…