A strong team on unfamiliar ground
The company ran its applications on VMs. The team was strong in AI and research, but had no experience with cloud or DevOps. Every deploy depended on manual work, and the environment didn't scale on its own.
From the start, it was clear the job had two parts: build the infrastructure, and leave the team able to run it without me.
Building
I designed and built the infrastructure from scratch on AWS. The applications moved into containers on ECS, the images into ECR, and the database became managed on RDS. I set up networking with VPC and security groups, permissions with IAM roles, the load balancer and centralized logs in CloudWatch.
All of the infrastructure was written as Terraform modules and delivered in the client's own GitHub, wired into the deployment pipeline. Nothing was left depending on an account of mine or a manual step only I knew.
Handing it to the people who will run it
Since the team wasn't an infrastructure team, the delivery included an operations guide and the infrastructure documentation: how to deploy, how to scale, where to look when something goes wrong. The goal was for them to run and evolve the environment without depending on the consultancy.
What changed
In about five weeks, the company moved off VMs to a containerized environment, with automated deploys and controlled scaling, and with a team able to look after it.
What I take from it
With a client that has no DevOps knowledge, documentation is part of the product. Infrastructure that only I understand becomes a single point of failure, just in the shape of a person.