MapR has announced six new data science service offerings to help customers gain immediate value from machine learning (ML) and artificial intelligence (AI).
According to MapR, the new offerings address the fact that AI and ML can be complex, as well as that organizations don’t always have the capacity to execute on AI and ML ideas, and those that do, may not be able to bring those ideas to production.
The six new MapR data science lifecycle service offerings include an AI/ML Hack-a-thon offering in which the MapR Data Science team works with the organization to identify a business use case and prototype a solution to deliver a real ML and AI solution that the organization will continue to improve and maintain over time.
In addition, MapR is offering the Data Science Refinery Accelerator, through which an expert will guide customers through installation, best practices, and baseline models to ensure maximum production success.
Addressing the possibility of “unknown” gaps that could allow evolving hackers to gain entry or inflict damage, the MapR cybersecurity data science offering orchestrates a real-time pipeline of logs (e.g., application logs, transaction logs, etc.) and trains models based on the unique signature of network sources and traffic. Ultimately, the organization receives a visual, UI-based assessment showing suspicious activity, allowing internal security experts to review and escalate threats in real-time.
Intended for organizations that are further along in their ML/AI journey, a new model deployment offering maximizes a model’s value by uploading the modeling process to the MapR Data Platform. The solution is then poised to take advantage of all the organization’s data, utilize every ML library, and deliver results that will scale and improve with the business.
An AI Enablement offering combines the MapR ML framework with Streaming events to deploy an AI engine that will begin to find new opportunities for optimization through a continuous learning and feedback loop. The team uses ML to bring order to the chaotic nature of a system’s behavior (e.g., a person, a car, a pipeline, etc.), then applies reinforcement learning to teach the system to adapt to identify and assess unusual cases to achieve generalization.
And finally, the ML Rendezvous Orchestration addresses the issues of ML models degrading over time. In many cases, MapR says, the arrival of performance results lags behind the next model deployment. Designed for mature ML processes, this offering enables organizations to monitor their ML workflows for events that might impact their accuracy in lieu of performance data. Having detected those impacts, the business will be able to make more informed decisions, such as when the current model should be replaced.
In addition, MapR announced the release of a new book, AI and Analytics in Production, by Ted Dunning, Ph.D., board member of the Apache Software Foundation, and chief application architect at MapR; and Ellen Friedman, Ph.D., principal technologist at MapR, and Apache committer. Published by O’Reilly Media, the book examines best practices for maximizing the value of data-driven applications in production.
For more information about MapR, visit mapr.com.