FAQ

When should I build an MLOps platform?

It’s time to standardize on an MLOps platform when you notice that your engineers are spending more time on researching and seeking approvals for tooling options than they are on building ML models.

Can I integrate other ML tools into the MLOps platform?

Yes. You can integrate non-AWS tools into the platform. While SageMaker AI Studio is at the core of the MLOps platform, you can still integrate other products with the SageMaker AI Studio suite of services.

How can my organization simplify governance requirements to accelerate innovation?

As part of the use case candidates that you select to prove the build of your MLOps platform, ensure that the use cases are of sufficient complexity, require various data classifications, and require big data volumes. By doing this, not only do you prove the platform capability, but you do the heavy lifting from a governance perspective as part of your initial platform release. If you can do this, then teams that adopt the MLOps platform as part of the rollout will have a lighter governance load, as they use a platform that has already met the governance requirements for complex use cases.

Which team do I need in place for building an MLOps platform?

A robust MLOps foundation, which clearly defines the interaction among multiple personas and technologies, can increase time to value, reduce cost, and enable data scientists to focus on innovation. Having the right team can be the difference between failure and success for MLOps platform development. Due to the nature of MLOps, many roles need to be involved, such as data scientists, ML engineers, DevOps professionals, data owners, IT owners, business analysts, and product owners. Make sure that all your stakeholders are interacting in a cross-functional team to ensure the best outcomes for your MLOps platform.

How can I start on my MLOps journey?

You can start by creating a secure experimentation environment where data scientists receive a snapshot of data. The data scientists can use SageMaker AI to experiment and ultimately prove that ML can solve a specific business problem.

Should an MLOps transformation be driven by a top-down or bottom-up approach in an organization?

While bottom-up approaches can be successful, support from leadership is essential for the success of MLOps platform development. With a top-down approach, you can ensure faster standardization of the developed solution, reduce costs, and achieve higher scalability and reusability between the models developed by different teams in your organization.