The short version

I run the internal platform four product teams ship on — provisioning, delivery, reliability and secrets — and I built the AI layer that helps operate it without giving an agent production access.

What I run Four product teams ship onto a platform I own, made of four layers: provisioning with Terraform across fifteen AWS accounts, delivery through a reviewed pipeline, observability against a 99.9 percent availability target, and secrets held in AWS Secrets Manager with one scoped role per app. Above it sits an AI operations layer that can read production and propose changes as pull requests, never apply them. The platform serves fifty services across three environments on EKS. 4 product teams AI ops layer · reads, proposes PRs provisioning — Terraform, 15 AWS accounts delivery — reviewed pipeline, 15+ deploys/day observability — 99.9% SLO, burn-rate alerts secrets — Secrets Manager, one role per app 50 services 3 environments the agent can read production and open a PR — it cannot apply anything
Everything above runs on EKS. The four layers are four project write-ups.Read any of them →

99.9% SLO held since early 2024 · 2 days → 1 hour to stand up an environment · triage 22 → 8 min

Resume (PDF) LinkedIn liana.gizatulina@gmail.com Tampa, FL · remote