Overview
Ongoing AWS infrastructure and operations for CartStack, a cart-abandonment and customer-recovery platform whose embedded JavaScript watches cart activity across thousands of customer sites. The relationship has run continuously since 2018, which changes what the work looks like: not a project with a handoff date, but a sequence of them.
The through-line is keeping a deliberately lean stack lean while the business grows. Early work was AWS administration and cost discipline — security group hygiene, database cleanup and backups, savings plans and spot usage. When CartStack white-labeled its platform for the Brazilian market, I split the infrastructure into a separate AWS account in a Brazilian region and ran the switchover of live customer sites, because latency to local users was the constraint that mattered.
The largest single piece was deployment modernization: containerizing the PHP application with Docker and moving to Git-based deployments on ECS via GitHub Actions and ECR, with EC2 autoscaling and the whole environment expressed in Terraform. The point was organizational, not technical — the setup a single developer can hold in their head becomes a liability the day a second developer joins, so the system came with diagrams, documentation, and training rather than just working infrastructure.
A recurring part of the role is independent technical judgment: helping a small team separate a vendor’s sales recommendation from what their architecture actually needs.