Microsoft Foundry

In general it is a very useful tool. The platform provides a centralised environment that allows us to build, test and deploy AI using foundation and custom models. This is very useful when you want to integrate the models in enterprise infrastructure. Documentation changes frequently, and this is the only hard thing to manage.

API is working as expected, also integration with Azure credentials sdk and other tools made our development much faster and operative quickly. New models are included frequently which is very important for the current evolution of AI. As a third point i can say that is great that it includes multiple models, image, text, speech and all you need in one single resource in Azure.

Seamless integration with Azure Ecosystem, which really helps us to accelerate and make the business streamlines stable with AI project development. Effective edge deployment. Lastly, it is highly effective and suitable for large-scale enterprises.

All the updates and changes have not been documented: as the technology is changing too fast and the models need to be improved, they hold a lot of legacy code and documentation that is becoming obsolete faster. Pricing will scale fast depending on the selected model so the necessity of using guardrails is mandatory to enable optimal solutions.

As mentioned in the discussion, during the implementation in our platform we faced that some of the resources that we needed to create were not yet available in the azure provider for terraform. We discussed with the support teams and we finally had to deploy it using scripts, which is not a best practice and duplicated the work we have to do in the future. As a user I think the UI can be improved, it looks very poor and not responsive enough.

Not beginner friendly, since it will require a lot of time to train. Limited community support, which means that users may encounter such time to troubleshoot if there is in need.