Get ready, Snowflakes: Azure AI is coming for you with one click

Microsoft is looking to make its Azure cloud the place for enterprises to run their AI and machine learning workloads.

"Data scientists from organizations that have adopted Snowflake as their data warehouse solution can now explore Azure ML capabilities without relying on third-party libraries or engaging data engineering teams," as Amar Badal, senior manager for Azure Machine Learning, put it in an announcement this week.

He added that thanks to native integration between Snowflake and Azure Machine Learning, those data scientists "can import their data from Snowflake to Azure ML with a single command and kick-start their machine learning projects."

The connectivity between the two platforms means data scientists can save time moving their data to Azure – either on a schedule or on demand – and then trace where that data's movements, Redmond claimed.

• Microsoft rains more machine learning on Azure cloud
• Microsoft injects AI search into Bing, Edge, Skype apps
• Microsoft's Copilot AI to pervade the whole 365 suite
• Microsoft would rather spend money on AI than give workers a raise

The announcement comes a week after Microsoft bulked up Azure Machine Learning at the Build 2023 conference with an expanded partnership with Nvidia and unveiled the public preview of foundation models in the cloud service.

It's clear Microsoft is making moves to strengthen its AI training capabilities, although how well that will work remains to be seen.

The software giant's latest moves include introducing tools that will enable organizations to more easily import information from data repositories that aren't part of the Azure platform – think database company Snowflake and Amazon Web Service's S3 service – into Azure Machine Learning for AI training. At the same time, Redmond is enhancing the tool suite used to track and manage AI training jobs in Azure Machine Learning.

As clouds offer elastic access to hardware tuned to AI workloads, which is often not the sort of kit affordable or comfortable to run on-prem, it's felt that enterprises will see hyperscalers as the cost-efficient way to embrace the rapidly accelerating AI trend being driven by large-language models (LLMs) and generative AI applications like ChatGPT.

Microsoft, like AWS, Google Cloud, and other providers, therefore wants to be the go-to cloud when it comes to AI workloads. Armed with the generative AI products it's got from the billions it's invested in OpenAI, Redmond has been on a months-long sprint to push machine learning into every part of its software portfolio.

This new initiative targets users of Snowflake's cloud-based data warehouse.

In conjunction with this, Microsoft is putting lifecycle management capabilities in public preview as a way to manage imported datasets in an Azure Machine Learning datastore, or what's called a HOBO (hosted on behalf of). It's a feature for data imported via the CLI and SDK.

"On choosing HOBO datastore as their preferred destination to import data, one gets the capability of lifecycle management or as we call it 'auto delete settings' on the imported data assets," Badal wrote. "A policy to automatically delete an imported data asset if unused for 30 days by any job is set on every imported data asset in the AzureML managed datastore."

The new tracking tools, which also are in public preview, are aimed at enabling enterprises to manage their training tasks. Included among them is a customizable dashboard view that pulls together chart visualizations of resource use, evaluation, and metrics for the jobs to get a more detailed overview of projects.

There also is a customizable list of training jobs that displays names of the tasks being worked on. Data scientists also can select and reorder columns, filters jobs via a range of criteria, and run batch actions on jobs.

Other new feature the ability to compared metrics and images from training projects, to add markdowns for notes, and create and save custom views.

Microsoft's clearly all-in on AI. ®