An infrastructure architect's opinionated take on what AI workloads actually need — GPU sizing, storage, networking, virtualisation, power and governance.
The full arc of a customer platform from presales proposal to a live, operated system — the hand-off chasm, decision records, and what production really means.
Hard-won lessons from running a real Docker homelab — compose-as-code, networking, Caddy, monitoring, updates, backups and the mistakes I actually made.
How I built an automated Microsoft 365 health assessment with Graph, n8n and a local LLM that turns raw tenant findings into a prioritised report.
How I design Citrix in 2026 — CVAD versus DaaS, hybrid control planes, identity, and honest advice on when not to use Citrix at all.
A first-person account of running large language models locally — GPU and VRAM choices, quantisation, Ollama, model selection by job, and where local inference beats hosted.
How I designed and evolved Atlas, my own local AI assistant, on Ollama, Open WebUI and n8n, with a Git-backed knowledge base and real tool-calling.
How AI changes solution architecture and technical consultancy — what it commoditises, what stays human, and why credibility now comes from the lab.
Why a home lab teaches engineering judgement that vendor labs and certifications cannot — owning the whole stack, breaking it, and fixing it yourself.
A living now-page and changelog of my current projects, experiments and deliberate non-goals across the homelab, AI work and presales practice.