I am responsible for setting/creating infra for the company and I am trying to introduce IaC approaches at the company. I am looking for ways and recommendations on how should I structure the terraform repository.

Infrastructure requirements can be categorized into two requirements, project specific and ad hoc requests. Example of project specific requirements could be something like, I need to set up a data platform for the company, a data platform would comprise of multiple infra components, like it would need glue, redshift database, roles to run glue jobs etc etc. Example of adhoc requests is like creating a new ec2 instance for small PoC or sharing access to a consultant for limited time etc etc.

^This is generally the type of requests I get, so based on this I am planning to implement this following directory structure.

- applications
  - data-platform
    - glue_role.tf
    - glue.tf
    - redshift.tf
  - some-other-application/
  - ad-hoc
    - ec2.tf
  - user-management
    - roles.tf
    - user-groups.tf
    - policies.tf
- modules
  - iam_role
    - main.tf
    - variables.tf
    - outputs.tf
  - ec2
    - main.tf
    - variables.tf
    - outputs.tf
main.tf

I also have to maintain three environments, dev, stage and prod. For environments I am thinking to use workspaces, so if I want to apply to prod, I should switch to prod workspace.

The main.tf file in the root, acts as an entrypoint for terraform command and it imports all other modules and directories (i.e. it imports data-platform).

In your experience, does this approach sounds correct and scalable? I will have other people also working on this, so I am also using s3 as a backend. Is it recommended to have separate backend per application?