This is just a quick reference for a complete article I published on Medium about a journey of migrating dozens of CI/CD pipelines from a internal Jenkins instance to GitHub Actions. Details can be found here.
On this process we felt like some of the pipelines should be executed on a custom infrastructure. This approach let us customize hardware needs (such as GPU, memory, CPUs) and access to private services inside a internal AWS VPC, adding another layer of security to our CI/CD workflows. This was the most ambitious part of the solution, taking leverage of ECS clusters and autoscaling groups to offer a scalable platform capable to fulfill our needs. This part is well documented on this repository and had another nice output: a article we wrote in partnership with AWS engineers. This last link is in portuguese and Spanish only. Enjoy! :)