Six projects, and what went wrong in each.

Every page covers the same five things: what it is, what I owned, what changed, how it works, and what went wrong. The last one is what most portfolios leave out, so it's where I'd start if you only read one section.

Agents
Reliability
Security

read / write split AI incident triage, without giving the agent production access An assistant that investigates incidents across 50 services and proposes fixes as pull requests — holding no cluster credentials and no permission to merge. 4 min read
30 repositories A second reviewer on every pull request An AI first pass that checks each diff against its ticket's acceptance criteria — advisory by design, never a merge gate. 4 min read
MTTR 45 → 15 min Giving a platform a reliability number it could defend A 99.9% availability SLO, an error budget people could actually see, and alerts worth waking up for. 5 min read
2 days → under an hour Making environments something teams create themselves Reusable Terraform modules across 15+ AWS accounts — a two-day manual runbook turned into a self-service build under an hour. 4 min read
zero code changes Making Secrets Manager the source of truth for live applications A migration under production traffic — a scoped IAM role per app, and not one line of application code changed. 5 min read
keyless Taking the long-lived AWS keys away Replacing static credentials in pipelines and workloads with federated identity — 120+ standing access keys eliminated. 3 min read

liana.gizatulina@gmail.comTampa, FL · remote-first