Design an ML platform for twenty teams

Twenty product teams currently train and deploy models however they like: some in notebooks on laptops, some in ad-hoc SageMaker jobs, two with a real CI pipeline. There are around 60 models in production and nobody can produce a list of them.

You have been asked to build a central platform. You have four engineers and twelve months. The teams do not report to you and cannot be compelled to adopt anything.

The stated goals: know what is in production, make deployments reproducible, and reduce time-to-production. Nobody has asked for the platform to run the training.

Design this — including your adoption strategy, since you cannot mandate it. Tell me what you build first, what you deliberately do not build, and how you would know in six months whether it is working.