In release 7.2, we have opened up DataRobot for data science experts who love to code via Composable ML and code-centric data preparation and pipelines. MLOps adds Continuous AI to keep production models at peak performance, bias monitoring to keep models fair, and new Decision Intelligence flows that let you apply business rules to each and every prediction.
Keep reading to understand more about our most exciting release yet.
Powerful Tools for Data Science Experts Who Love to Code
Your Expertise Extended with Our World-Class Automation. To build best-in-class models, data scientists need to constantly experiment with data and algorithms. In reality, data scientists spend much of their time on repetitive or mundane tasks, like writing code for feature transformations or model operationalization, leaving them with little time to actually experiment.
New in release 7.2, DataRobot Composable ML provides customizable blueprints containing reusable building blocks that allow experts to save time on mundane coding tasks and focus on experimentation and other high-value activities. You can define your unique best-in-class machine learning algorithm, then combine it with built-in DataRobot’s capabilities to automate repetitive and laborious tasks, resulting in a dramatic boost in productivity.
Prepare and Publish Data Flows with Code. In release 7.2, we now offer code-centric data pipelines for data scientists, data engineers, and SQL savvy analysts. This new capability enables you to create a workspace in the AI Catalog, then connect to your data sources and build self-contained data preparation modules using the coding languages you know and love.
You can organize these code modules into a pipeline or graph format via an easy-to-use flow editor that can be scheduled and run how you choose. Code-centric data pipelines give your coding experts the ultimate flexibility to quickly create new datasets that can be used in any machine learning project.
Notebook Powered Data Science. DataRobot now has an integrated notebook solution with its recent acquisition of Zepl. Sign up for your free trial and get pre-built notebooks with all the code you need to connect to DataRobot using Python or R. Use Notebooks to explore data sets and do ad hoc analysis in Snowflake, Amazon S3, or other cloud data sources. Visualize your findings with our built-in Plotly editor or use the data visualization library of your choice. Easily share results with your business users with our “publish” features. Our cloud-hosted notebooks require zero maintenance from you or your team. Get all the goodness of notebooks without the headaches with DataRobot Notebooks fully-managed cloud solution.
External Prediction Insights
Out of the Box Explainability For Any External Model. New in Release 7.2, external prediction insights allow you to bring the output of any model you already created outside of DataRobot, into our machine learning development environment. Once uploaded, you can compare your existing model with any best-in-class model generated on the AutoML leaderboard. Additionally, you can evaluate your external model using our out-of-the-box visual insights and explainable AI tools, including the Lift Chart, ROC Curve, Bias and Fairness insights, and more.
Model Quality for your Production Models
Keep Production Models at Peak Performance. With Continuous AI, you can create multiple MLOps strategies to refresh your production models. You can base them on your own schedule or set them to be triggered when accuracy drops or data drift occurs. Continuous AI also uses DataRobot’s world-class AutoML capabilities to automatically create and recommend new challenger models for you. When combined, these strategies ensure that you always have the most up-to-date models generating the most accurate and timely predictions possible.
Bias and Monitoring in Production
Proactively Monitor Your Production Models for Bias. Bias monitoring is new in DataRobot Release 7.2 of MLOps. Machine learning models that do not exhibit bias at training time can become biased after they are deployed. As human behaviour changes so will data we generate, which could cause your model to become unfair. MLOps now allows you to monitor your production models using the fairness metric you selected when you trained the model, such as proportional parity. You can also apply thresholds to trigger alerts if your model falls below the benchmarks you set. If bias is detected, the data drift insight allows you to understand how your data is changing over time and help you diagnose the source of bias.
AI-powered Decision Making
No-Code AI Apps
Making AI Accessible to Front-Line Decision Makers. DataRobot’s No-Code AI Apps is now generally available. It allows you to quickly turn any model into an AI application without requiring any coding. Use our pre-built templates and drag-and-drop widgets to quickly configure your apps. Just pick the model features and data visualizations you want your app consumers to use in their decision-making process. This makes it much easier for business users and information workers to leverage predictions generated by their models and make more informed AI-driven decisions.
Scaling Your AI Projects
Pathfinder Solution Accelerators
Quickly Get Started with the Most Popular AI Use Cases. DataRobot Pathfinder Solution Accelerators are an application-first AI exploration experience. Pathfinder Solution Accelerators provide a library of common AI use cases and a variety of tools and integrations to jumpstart your path to value with AI. Like an AI marketplace, Pathfinder Solution Accelerators make it easy to find, preview, and get started with popular AI use cases fast. In just a few clicks, you can go from theory to training your own model on DataRobot’s world-class AI platform in a completely seamless and guided experience.
Enterprise Platform Readiness
AI Cloud Platform
Even More Enterprise-Ready. In release 7.2, we have worked hard on our architecture and IT administration capabilities to make the platform even more enterprise-ready. DataRobot administrators can now set multiple roles per user and permanently delete users and their associated projects and deployment data. DataRobot modeling workers now run on Kubernetes for managed cloud offerings to improve the overall reliability, workload management, and auto-scaling capabilities of the DataRobot platform (support for non-premise and VPC offerings in the works).
More Coming Soon. Stay Tuned!
Other new features include Nowcasting for Automated Time Series, anomaly detection for Visual AI, new public documentation, and many other critical enhancements.
These are just some of the major highlights of the DataRobot 7.2 Release. For a complete list of new and enhanced features, please visit the DataRobot Documentation Release Center or join the conversation in the DataRobot Community.