In Release 7.3, we are thrilled to announce expanded capabilities and features for all users to enable AI-driven decisions across all lines of business within a single platform.
Composable ML and code-centric pipelines are now available for SaaS users. We also took modeling to the next level. Now you can leverage new, out-of-the-box capabilities, such as multimodal clustering, segmented modeling, unlimited multiclass, and multilabel classification, and combine them with data you have to solve a diverse set of challenges.
As part of the MLOps product, automatic compliance documentation helps you accelerate your preparation for regulatory reviews for any model. In addition, with built-in Challenger Models Comparison, you can review and compare model behavior for the champion or challenger model before deciding on a model replacement.
In addition, Release 7.3 extends capabilities for our No Code App Builder. Now, the easy-to-use, customizable app builder supports time series models, allowing you to build more AI-driven apps for different complex use cases with no code.
These are just a few headlines. Let’s dive in!
Code-Centric Data Science
Augmenting Your Expertise with World-Class Automation. We are happy to share that DataRobot Composable ML is available in the 7.3 Release for SaaS users. Composable ML provides customizable blueprints containing reusable building blocks so that data science experts can save time on managing libraries, editing code, and prepping routine model operations tasks and focus instead on experimenting with data, working with advanced algorithms, and other high-value data science activities. DataRobot can automatically generate and train blueprints to bring good options to the surface, which a data scientist can marry up with custom code to create new tasks and incorporate any modeling logic you want. Share the components you create, and all users in your organization will benefit from your work. Focus on the task and the code, and seamlessly integrate with a wide range of DataRobot capabilities with no extra DevOps work: Leaderboard, Insights, Compliance docs, Deployments, Monitoring and Governance, and more.
Code-Centric Data Pipelines
Prepare New Datasets Without Worrying about the Scale. Code-centric data pipelines are now available for all SaaS cloud users. DataRobot pipelines allow data scientists, data engineers, and SQL experts to quickly build and orchestrate sophisticated data pipelines and reusable data preparation modules, using coding languages they know and love.
Expanded Support for Diverse Use Cases
Deliver Actionable Insights to Your Business in No Time. Multimodal Clustering provides users with a one-click, one line-of-code experience to build and deploy clustering models on any data (e.g., numeric, categorical, text, image, and geospatial). You can easily understand, name, and explain each cluster for any model with the new Cluster Insights visualization.
Multimodal clustering expands the number of use cases dramatically, from topic modeling, market segmentation, new product development and selecting test markets, and much more.
Diverse Languages and Models. Do More with Text AI. Leverage text data you have, in any language, to streamline operations, deliver insights, and bring more value from AI to your organization. We embedded the best language-agnostic data science practices into simple and powerful, out-of-the-box capabilities that help take your text models to the next level. Drag and drop text data into DataRobot AI Cloud Platform to build, deploy, monitor, and manage text-based models with just a few clicks. With Text AI, you have full control over tokenization with a diverse set of options for stemming, lemmatization, embeddings, and more. Find more details in our recent blog.
With Composable ML, you can bring your own text featurizer or text modeling algorithm to the AI Cloud Platform and build your advanced model. You can also put text into a bigger context. By combining text with numeric, categorical, image, time series, and geospatial features, you can build a better model and understand multiple data dimensions simultaneously.
Visual AI Anomaly Detection
Maximize Business Value from Your Visual Data. With Visual AI Anomaly Detection, you can now address more use cases out-of-the-box, with one click and one line of code end-to-end. You can leverage Embeddings, Activation Maps, and Prediction Explanations, literally seeing what your model is paying attention to. It’s the perfect way to get valuable insights and tell a convincing story about your modeling choices and the individual anomaly scores. And, like with anything else on DataRobot AI Cloud, you can combine images with other data types to solve a diverse set of challenges and unlock the value of AI without additional effort.
Time Series Segmented Modeling
Effortlessly Build, Manage, and Deploy Highly Accurate Forecasts Across Hundreds of Segments. Companies trying to forecast multiple series are often faced with a choice: roll all your time series up into one big series and fit “one model to rule them all,” or manually build a separate model for each series. In today’s volatile environment, neither option is acceptable.
In 7.3, DataRobot Time Series, we make it effortlessly easy for you to create, deploy and manage highly accurate forecasts across hundreds of segments and thousands of items with time-aware segmented modeling. DataRobot will automatically detect when you will benefit from segmented forecast modeling. Then, simply define segments based on business logic to split your series up. If you are a retailer, for example, you can segment your data by product type, and DataRobot will automatically surface fine-tuned accurate models for each segment. You can quickly analyze results at the global level, or dive deep into each segment. Once ready for deployment, intelligent model roll-up neatly packages segments into a single project to streamline management, and fast-track deployment of hundreds of models. Time Series segmented modeling empowers you to dramatically improve your forecasts and deliver significant value back to the business with a fraction of the effort.
Enhanced Performance Monitoring, Compliance and Regulatory Capabilities
Be Regulatory Review-Ready with an Automated Compliance Documentation. In the 7.3 Release, Automatic Compliance Documentation is available for all users, allowing you to automatically generate compliance documentation from the MLOps Model Registry for any model, including custom models. DataRobot MLOps is a centralized hub for any model in production, regardless of how and where they were deployed. As a result, you can bring in any model and have full control over your production AI.
Challenger Model Comparison
Optimize Model Performance after Deployment with Confidence. To feel comfortable replacing your deployed model with another, you need to have confidence that another model is performing better. In the 7.3 Release, we have expanded the Challenger framework to include additional insights focused on direct comparisons between the champion and challenger model. These model insights will help you understand and compare the behavior of two different models before making an important decision on model replacement. You can generate insights to evaluate your model (Accuracy Metrics, Dual Lift, Lift, ROC, or Prediction Difference) and choose the top-performing model with confidence.
Better, Faster Decisions with Decision Intelligence
Time Series Support for No Code App Builder
Create an AI App. No Code Required. Looking for a way to empower your frontline decision-makers with actionable forecasts? In the 7.3 Release, we added time series support for our No Code App Builder. With simple drag-and-drop widgets, you can build a fully customized AI app on top of a time series model, deployed in a matter of minutes with no coding required. For example, compare forecasted versus actual values over time, and then dig into the reasons driving each forecast.
Decision Flows Enhancements
Make Decisions with Speed and Scale. Now you can deploy your decision flows into production via your API and make thousands of decisions in real time. All deployed decision flows are accessible from the Decisions Inventory tab, where you can monitor the number of decisions made, the health of the flow, and activity status. Decision flows are able to return numeric or categorical outputs based on combinations of multiple models in a single decision flow. You can also copy or share decision flows so that you can collaborate with your team.
These are just some of the major highlights of the DataRobot 7.3 Release. For a complete list of new and enhanced features, please visit the DataRobot Documentation Release Center or check demo recordings. Join the conversation and ask questions about the release in the DataRobot Community.