When I'm not pushing buttons, I'm trying to not set off my smoke alarm, burying my nose in a fantasy novel, and performing my duties as a cat dad. What's left of my time goes toward endlessly tinkering with my homelab when I get that "is Plex down?" text message.
An event-driven pipeline that handles 20M+ transactions an hour on finite hardware. Autoscaling kicks in before the bursts land. Pre-warmed caches keep aggregations inside the 1-hour SLA.
Rolled out a shared design system to 50+ apps across 30+ teams, on a legacy financial platform that's been around longer than most of us. 7 apps migrated in 6 months, 28 more in flight. Developer velocity up 25%, DX satisfaction up 50%.
A SAML-heavy auth backbone no one on the team fully understood. We rebuilt it piece by piece, without downtime. RAG surfaced what the old code actually did, and agent skills wrote the docs as we went. Context coverage went from under 20% to 85%. Onboarding dropped from 4 weeks to 1.
Multi-AZ Talos Kubernetes on Proxmox, declaratively managed with Ansible + ArgoCD. Same GitOps I use at work. The difference: when the power goes out, I'm the one with the flashlight.
Deny-by-default, allowlist-first. Six VLANs, VPN-only remote access, DMZ for anything that touches the public internet.
Two TrueNAS nodes running ZFS, with Cloud Sync + RSync replicating across availability zones. The cat videos are not allowed to disappear.
A local LLM watches my monitoring stack. It triages new alerts and tunes thresholds when the signal drifts. Ansible still runs Prometheus and Grafana. The LLM gets the 3 AM pages now.
Three Talos clusters, all managed from Git. ArgoCD does the syncing, LGTM does the watching. This site runs on it, so if the page loaded, the lab's alive.