What are people’s thoughts on Microsoft Foundry?

Just as Microsoft Fabric serves as a data platform, Foundry acts as a platform for AI agents. I see a lot of potential here, particularly due to its seamless integration with the broader Microsoft ecosystem and features like content filtering and usage monitoring.

For those of us focused on inference and building applications on our own Kubernetes setups, Foundry feels overwhelming. Before Foundry came along, things were a lot simpler. Microsoft's frequent updates are adding to the confusion, but I see their intent to cater to users not on Azure.

Microsoft Foundry is essentially a centralized platform for AI, combining various services like Cognitive Services, Document Intelligence, and more into one place. The main benefit is having everything available from a single dashboard, simplifying the learning curve since users don't have to get into the complexities of Azure. However, it can be tough to keep up with all the rapid changes, and some services only offer free tiers when set up as separate resources.

The rate limits on Foundry can get really frustrating. You can easily hit rate limits, making it tough to use many models, especially in terms of code execution. It feels like you're running into access issues constantly.

One major drawback I've found is that your projects are essentially locked in once you build them out. If you attempt to move to a different subscription, you're forced to either rebuild everything or scrap it entirely. I've been stuck with two tough choices: rebuilding everything in a new subscription or moving elsewhere since the value just isn't compelling anymore.