As machine learning becomes more mainstream across industries, it is also gaining popularity across diverse business functions, from marketing and sales to operations and finance. How an organization employs AI often determines whether each of these groups can successfully realize the technology’s full potential. These organizations look to unified machine learning platforms to manage the entire ML lifecycle and bring models into production faster and at scale. Machine learning operations (MLOps) plays a critical role by bringing together people, processes, and technology to automate ML-infused software delivery and deliver greater quality and consistency to users.
Azure Machine Learning helps you streamline the deployment and management of thousands of models across production environments, on-premises, in multi-cloud, and at the edge, using MLOps capabilities such as model registry, managed endpoints for batch and real-time inference, CI/CD pipelines, and experiment tracking and lineage.
One great example of a customer transforming their business with machine learning and MLOps is Brazilian-based consumer goods company, BRF. BRF adopted Azure Machine Learning to develop machine learning models that improve business outcomes and sales forecasts. They use MLOps best practices and capabilities to automate all phases of the machine learning lifecycle. Because of that, their analysts have more time to focus on strategic, value-add work rather than manual model management.
“We’re scaling with automated machine learning in Azure and MLOps capabilities in Azure Machine Learning so that our 15 analysts can focus on more strategic tasks instead of the mechanics of merging spreadsheets and running analyses.”- Alexandre Biazin, Technology Executive Manager, BRF.
To help organizations uncover some of the challenges and nuances of choosing an MLOps platform, the analyst firm GigaOm recently published their Enterprise Readiness of Cloud MLOps: A GigaOm Benchmark Report. This report assesses leading MLOps platforms’ capabilities, features, ease of use, and documentation. GigaOm partnered with Microsoft, the report’s sponsor, to select competitive platforms that offer comparable features and capabilities to address enterprise organizations’ needs. GigaOm selected the test scenario, methodology, and configuration of the environments and published their methods step by step so that any organization can reproduce the tests during their own independent evaluation.
GigaOm gave Azure Machine Learning the highest overall assessment score in enterprise capabilities and time to value.
GigaOm specifically recognized Azure Machine Learning as the easiest platform to set up a test, orchestrate the model and data, and set up security. They gave Azure Machine Learning the highest scores in all 5 evaluation metrics. Let’s take a closer look at the capabilities where Azure Machine Learning emerged as a leader:
- Ease of setup and use – Azure Machine Learning offers an intuitive point-and-click interface in workspace. This enables you to set up compute instances and a production-grade cluster, as well as development environment with Notebooks, Visual Studio Code, drag-and-drop Designer, and automated ML.
- MLOps workflow – Azure Machine Learning scored full marks on all data, model, and pipeline orchestration. The Designer capability provides drag-and-drop data orchestration, including data transformation, normalization, and descriptive statistics of columns of data. With model registration, you can store and version your models in Azure cloud, in your workspace. The model registry makes it easy to organize and keep track of your trained models. Azure Machine Learning also offers Azure ML Pipelines through SDK, CLI, and Designer that stitches together various ML phases. Native integrations with a popular open-source framework MLflow can help you easily track an experiment’s run metrics and store model artifacts in your Azure Machine Learning workspace.
- Security – This is a core investment area for Azure Machine Learning, which received full marks across Network Security, User Security, and Data Security. The platform offers fully-realized security features by isolating workspaces and training environments with virtual networks and private endpoint. For user security, Azure Active Directory can assign RBAC roles to users, groups, service principals, or managed identities to grant or deny access to resources and operations. Additionally, Azure Machine Learning integrates well with a variety of Azure data services and enables data encryption.
- Governance – GigaOm found that only Azure Machine Learning allows users to set up network and data protection policies that, for example, will make sure users cannot create workspaces with public IPs or without customer-managed keys. Azure Machine Learning also has a built-in monitoring capability with Azure Monitor that allows users to track key pipeline metrics and resource logs.
- Automation – Azure Machine Learning offers automated ML to perform automated experiments that help users find the best model based on a success metric of their choosing. Azure also supports automated versioning for code and application orchestration through GitHub Actions and Azure DevOps integrations that can be leveraged for a release pipeline with stages and approvals.
Read the Enterprise Readiness of Cloud MLOps: A GigaOm Benchmark Report for more details and to learn how
a unified MLOps platform like Azure Machine Learning can help you put your innovative models into production faster and at scale.
More Information
- Learn more about Azure Machine Learning
- Watch the latest MLOps breakout session from Build 2022: Scaling responsible MLOps with Azure Machine Learning
- Try Azure Machine Learning for free