We built the tool we wished existed at 3 AM.
In 2021, Thomas Rivera was Staff SRE at a Miami-based logistics company running over 600 microservices on Kubernetes. Every bad deploy followed the same script: alert fires at 3 AM, SRE runs kubectl rollout undo half-awake, incident report gets filed, runbook gets a new paragraph. He spent six months building an internal canary controller that caught bad deploys before they hit users. It worked. In 2023, he left to build it properly.
Making zero-downtime deployments the default, not the exception, for every engineering team.
Most engineering teams solve the bad-deploy problem the same way: custom automation scripts, ad hoc runbooks, or a dedicated platform engineer whose entire job is watching Grafana at night. Every company rebuilds the same canary logic from scratch. We find that absurd.
Zero-downtime deployment should not require a dedicated team. The logic — route a slice of traffic, query Prometheus, evaluate the burn rate, promote or roll back — is well-understood. What was missing was a controller that made it declarative, auditable, and reliable enough to trust without a human watching.
Kubestead is not a CI/CD pipeline, not an observability platform, and not a replacement for Argo CD. It owns one thing: the 90-second window between kubectl apply and "users notice." That constraint is intentional.
The people building Kubestead
Staff SRE at a logistics company in Miami from 2019 to 2023, responsible for deployment reliability across 600+ microservices. Built the internal canary rollback controller that became the basis for Kubestead's core engine.
Kubernetes upstream contributor (sig-apps) from 2020 to 2022, previously senior engineer at a distributed systems startup in Austin. Designed Kubestead's controller reconciliation loop and the AnalysisRun state machine that tracks every rollout verdict.
Built the observability and deploy tooling at a fintech company as it scaled from 20 to 200 microservices between 2018 and 2022. At Kubestead she owns the metrics integration layer — the adapters for Prometheus, Datadog, and the HTTP-based custom backend API.
Five years at a developer tooling company working directly with platform engineering teams at growing-stage software companies. At Kubestead he owns onboarding, the documentation structure, and the structured feedback loop from early adopters — which has driven eight of the last twelve feature decisions.
Company facts
Miami, FL
Angel round, February 2026
Microservices per team on average