Sepehr

Background

Sepehr has been writing software since he was 13, and has worked professionally as a software engineer since 2009. He is currently a Head of Engineering, leading an organisation of more than 20 engineers across backend, frontend, and infrastructure disciplines.

Smart home systems came next, about ten years ago, starting with a handful of smart plugs and a Philips Hue bridge. That grew into a full self-hosted homelab that now runs continuously on dedicated hardware: Home Assistant as the automation hub, Zigbee2MQTT bridging a mesh of Zigbee sensors and switches via a USB dongle, Mosquitto MQTT as the message broker, a Ring doorbell integrated locally via ring-mqtt, Frigate NVR running on an OpenVINO-accelerated GPU for local video analytics, and face recognition powered by Double-Take and DeepStack — no cloud required for any of it.

The media side of the lab runs a full *arr stack (Sonarr, Radarr, Prowlarr, Bazarr) alongside Jellyfin. Infrastructure is managed through a Cloudflare Tunnel and Traefik reverse proxy, with Prometheus and Grafana providing real-time observability. Pi-hole handles DNS-level ad blocking across the whole network. Everything is backed up nightly to S3. A Homer dashboard ties it all together.

This is not a weekend hobby setup — it is a production-grade system that has been running, failing, being debugged, and being improved for years. That experience is the foundation of every article on this site.

How I write

Every article on this site starts with a real question a UK smart home owner is likely to have. I research the topic using primary sources — manufacturer specifications, the official Home Assistant documentation, Which? consumer research, Ofgem publications, and gov.uk where relevant. I never cite Reddit threads or YouTube videos as evidence for factual claims.

Where a product or integration is relevant to my own setup, I test it directly and write from first-hand experience. An AI assistant helps me structure drafts and surface gaps; I then review every sentence, correct errors, and add first-hand detail the AI could not have known. Every factual claim is verified against a cited source before publication.

I accept no sponsored content, no free review units, and no PR relationships. Affiliate commissions help cover running costs but play no role in which products I recommend or how I score them. The editorial policy is on the editorial policy page.

My setup

This is the actual stack I run at home, not a theoretical one. Every component below is currently live and in daily use:

• Home Assistant — running on dedicated always-on hardware; the central hub for all automations
• Zigbee2MQTT + Mosquitto MQTT — a USB Zigbee dongle bridges a mesh of sensors, switches, and bulbs without any manufacturer cloud
• Ring doorbell via ring-mqtt— Ring integrated locally into Home Assistant over MQTT, no polling of Ring's cloud for automations
• Frigate NVR on OpenVINO GPU — local video analytics and object detection, entirely on-device; no video leaves the network
• Double-Take + DeepStack — face recognition pipeline for identifying people at the front door, running locally
• Full *arr media stack — Sonarr, Radarr, Prowlarr, Bazarr, and Jellyfin for self-hosted media management
• Cloudflare Tunnel + Traefik — secure remote access and reverse proxy without opening ports on the router
• Prometheus + Grafana — real-time monitoring and dashboards for all services
• Pi-hole — DNS-level ad and tracker blocking for the entire network
• Homer dashboard — self-hosted start page linking all services
• Nightly S3 backups — encrypted backups of all configuration and data