Azure Machine Learning and DataRobot Team Up to Accelerate Value of Generative and Predictive AI

The Azure Machine Learning team is thrilled to share a groundbreaking integration between Microsoft's Azure Machine Learning (AzureML) and DataRobot, a leader in Value-Driven AI, that comes as a result of the recently announced partnership. This integration brings together the power of Azure Machine Learning's capabilities with DataRobot's expertise in accelerating the full life cycle for building, deploying, and monitoring enterprise-grade AI solutions.

The DataRobot AI Platform uniquely brings both generative and predictive AI capabilities together in a unified, open, and end-to-end environment. With the new integration, the DataRobot AI Platform is now easy to run directly on Azure Kubernetes Service (AKS). Data scientists can now leverage the power and capabilities of Azure OpenAI Service in DataRobot Notebooks Code-Assist. They can also deploy models for real-time and batch use cases through Azure Machine Learning managed online endpoints, all while monitoring these deployments within DataRobot.

That means, for organizations that are customers of both DataRobot and AzureML, the new integration offers a streamlined workflow. Teams looking to create a model in DataRobot can take advantage of unmatched Generative AI assistance. This boosts productivity and performance and can be customized for specific business use cases. Once the model is ready, it can be deployed for inference on Azure Machine Learning. Using managed online endpoints, data scientists can easily test or swap models using blue/green deployment. Additionally, robust monitoring capability identifies stale models, preventing business risks.

Fast-Track Your AI Initiatives

Data scientists rely on Azure Machine Learning for its broad set of tools for building and deploying models, but may wish to accelerate important initiatives with the help of a collaborative AI platform and generative AI that can help them experiment and test models faster. ML engineers may desire ways of bringing additional observability, bias and custom metrics to their deployments, with the flexibility of a managed online endpoint. With the new integration, both model experimentation, and deployment become more streamlined for practitioners. It lets data scientists quickly iterate on models in DataRobot with the scale and data governance that comes with integration on AKS, deploy to a managed online endpoint with one click, and then use DataRobot to provide machine learning observability integrated with blue/green deployments whether the use is online or batch.

From Data to Deployment and Generative AI: What’s new in the integration

Data scientists can now build and train their models in DataRobot on top of Azure Machine Learning using their preferred authoring experience and frameworks. Once they are ready to get feedback or push the model to production, data scientists will deploy their models using the managed online endpoints feature in Azure Machine Learning. With the preview integration with Azure OpenAI, they can leverage cutting edge LLMs to assist with writing code to update their models or tackle new use cases quickly.

The integration enables seamless usage of the best from Azure and DataRobot in one workflow:

• Organize and prepare data using Azure Machine Learning pipelines from sources in Azure Data Lake Storage Gen2 and Microsoft Fabric.
  
• Link DataRobot with Azure data sources for seamless data integration.
  
• Experiment and build models on DataRobot, either through its hosted notebooks or via its user interface, all running on top of Azure Kubernetes Service..
  
• When using DataRobot's hosted notebooks for experimentation, you can utilize OpenAI's code assist feature. This provides conversational prompts that automatically generate Python code for tasks like data preparation, insight gathering, and optimization.
  
• Deploy trained models, using scoring code, to Azure Machine Learning managed online endpoints using blue/green and traffic shaping features
  
• Use DataRobot to automate model compliance documentation for both DataRobot and Azure models
  
• Monitor and manage the use case over time for drift, accuracy, bias and business KPIs in DataRobot
  

For example, in a customer churn scenario, a data scientist can work with business experts to organize the data using Microsoft Fabric and Azure Machine Learning, and define what churn means for a given set of customers. Then they can import it into DataRobot and use code-first or low-code options to get fast analysis of the business problem and build transparent and accurate models using the best approaches from deep learning, open-source and proprietary models, and more including out-of-the-box graphs to evaluate and get insights from the model . After finding the right model, then can collaborate with internal stakeholders like a model risk manager to validate this model and get approvals to deploy, before setting it up behind a managed endpoint. After testing and integration with the business process, such as building an Azure Machine Learning pipeline to connect the model to the businesses CRM system, they can operationalize it and monitor performance in real-time using DataRobot.

Get Started with the Azure Machine Learning and DataRobot integration

● Check our quickstart documentation

● Guided Videos:

● Use Azure Marketplace to purchase DataRobot, create an Azure Machine Learning resource, and start your journey

Continue reading...