Reliable platforms. Boring deploys.

I was the only infrastructure engineer at KnowNow Africa, a production SaaS company running on AWS and GCP. No escalation tier, no second engineer to hand off to at 2am. Every incident, every deploy, and every architecture decision was mine alone for four years. I cut release cycle time by 40 percent, built Prometheus and Grafana observability from nothing, and ran incident response and post-mortems as the only responder on call. That scope is why I read your BIBEs and canary work and recognize the exact shape of the problem, not because I've used every tool in your stack, but because I've carried the pager alone long enough to know what actually breaks.

A GitOps controller I wrote in Go, watching GitHub webhooks and reconciling Kubernetes manifests directly against the API rather than wrapping an existing tool. I built it to understand reconciliation logic from the inside. That's the same muscle your BIBEs (Buffer Isolated Build Environments) need: per-PR environments that come up clean and don't leave orphaned resources behind.

A terminal-native developer environment for deterministic replay and offline verification of API and LLM workflows. Designed to isolate failure modes locally without relying on recurring SaaS spend or flakiness, aligning with Buffer’s policy of using AI to remove operational toil for engineers.

Four years without an escalation tier means every deploy had to have a rollback path I could execute half-asleep. Git stays the source of truth, ArgoCD syncs it, and Prometheus-driven health checks decide whether a release holds or backs out. Argo Rollouts and KEDA would sharpen this further, canary steps instead of a hard cutover, queue-depth scaling instead of reactive CPU triggers, and they're what I'm learning right now ahead of talking to your team.

Empathy Built from Active Publishing

Buffer's platform succeeds when creators post without interruption. Alongside the infrastructure work, I run two developer communities with 280 and 208 members, have given 100 or more technical talks in five years including AWS Summit Johannesburg to 15,000 delegates, and write and record for an audience on dev.to and YouTube every week. That's four years of knowing what it feels like when a platform goes quiet on the people depending on it, so reliability isn't abstract to me, it's the difference between someone's audience showing up or not.