In October 2025 I left Mozilla, and I’ll be honest about where my head was: the market was rough, and I knew the biggest hole in my skill set was active development. I had spent years operating infrastructure at scale without being asked to write much code, and I’d let that slide for too long. It felt like a huge mountain to climb.

Eight months later I’m a DevOps engineer at Jasper AI, the team’s ArgoCD SME, and I’ve shipped a free Kubernetes training platform that strangers on the internet actually use. This post is about what happened in between — not because the story is special, but because the approach is repeatable.

Step one: build something real, in public

Instead of grinding LeetCode, I built the platform I’d want to operate professionally. My homelab went from a simple Docker setup to a multi-node K3s cluster with GitOps delivery through ArgoCD, Terraform and Ansible for provisioning, VXLAN-isolated networking, and a full Prometheus/Grafana/Loki observability stack.

The point wasn’t the hardware. The point was that every skill I felt shaky on — writing code, structuring Terraform, debugging Kubernetes from first principles — got exercised daily, with real consequences when I got it wrong.

Step two: use AI as a multiplier, not a crutch

I leaned hard on AI assistance to accelerate the learning, and I wrote about that honestly — including the time it helped me silently break my own platform for 24 hours. The lesson that stuck: AI raises the ceiling for engineers who stay in the loop and verify everything. It doesn’t replace the judgment you build by operating systems yourself.

Step three: the gap closes faster than you think

In January 2026 I joined Jasper AI. The homelab work transferred directly: within months I had taken platform ownership of our ArgoCD delivery pipeline, led a Datadog observability overhaul across multiple engineering teams, and handled an urgent supply-chain security remediation.

The development gap I was so worried about? Building K8s Learn — a Go-backed training platform with a browser terminal, running on my own cluster — answered that question for me.

If you’re in the October-2025 version of this story

Three things I’d tell anyone staring at the same mountain:

  1. Build the thing you want to be hired to run. A homelab with real GitOps, real monitoring, and real incidents teaches more than any course.
  2. Write it down publicly. Every post on this site doubled as interview material.
  3. The gap is smaller than it feels. Eight months of deliberate, hands-on work changed my trajectory completely.

If you’re working through something similar and want to talk platforms, GitOps, or career transitions — get in touch.