sizing cloud resources for an AI credit risk analysis application

Question

AI SaaS/CRM solution focused on enhancing b2b credit risk analysis. The customer is trying to size his solution in Azure cloud for investors
but isn’t clear on how much compute, networking, etc he’ll need and if any of that is included in the AI/ML framework solutions Azure cloud offers.

Answer 1

@J P Is the customer looking to use their own models on the Azure Machine Learning framework or service?
If Yes, then the compute requirements of the customer are not limited by the machine learning workspace they are going to use. The compute required to run the training and inference are infact the same compute instances that Azure offers as Azure compute or virtual machines. This can be scalable based on the kind of experiment or compute type chosen i.e virtual machine compute cluster or AKS or ACI. The only charge for using the machine learning service is the compute or storage or networking that is used for your experiments.

Most of the cognitive service related offerings are API based which are limited based on the pricing tier that is chosen. If a certain pricing tier falls short based on usage then the customer can always upgrade the tier and they are offered as pay as you go model. The compute that powers the APIs are highly available and scalable in the backend so higher TPS is supported as per the pricing tier chosen.