Dell Analytics Anywhere
Powered By BOSS AI
Enterprise Data Analytics and AI
Dell Analytics Anywhere empowers businesses and organizations to solve today’s biggest challenges through predictive analytics, furthering process automation, and accelerating digital transformation. At the core of Dell Analytics Anywhere is BOSS AI, a powerful data analytics and AI engine that can process any kind of data wherever that data resides. Through the use of Machine Learning and Deep Learning modeling of Big Data, your customers can build sophisticated algorithms to surface impactful insights or operationalize transformational technologies.
A key value of Analytics Anywhere is providing a single operating environment for all data analytics and AI development projects. Companies that have successfully deployed AI into production have done so through establishment of cross-functional teams, wherein the data science team works alongside domain experts, line-of-business leaders and others who will either support or benefit from implementation of the AI capability. With Analytics Anywhere, we provide a permissions-driven, project-oriented workspace for all data analysis and model development workflows. Powerful and immersive data visualizations, explainability tools and ‘click versus code’ workflow generation deliver operational simplicity necessary for cross-functional teams to collaborate.
AI/ML modeling is an iterative process. Some models deliver results immediately while others require rework and retooling of source data and model parameters. Traditionally, data scientists have utilized their own equipment, tools and techniques for model creation and training which encumbered versioning and repeatability of model iteration. The built in Asset Library solves this challenge being a single store for data queries, datasets and models with annotation providing a clear understanding of how the data assets were created and what was done with them.
One of the biggest technology barriers to AI success is inclusion of distributed data in Machine Learning model training and testing. Traditional modeling techniques typically start with copying data into a centralized system. Oftentimes this is prohibited due to compliance regulations, data ownership protections, network or other technology constraints. This results in projects failing due to not having access to the correct data or adequate sample sizes to generate accurate results. Federated Learning solves this challenge by distributing and operating models where the data resides, thereby circumventing the need to copy data into a centralized system.
Through BOSS AI, your customers will be able to perform both horizontal and vertical Federated Learning. In horizontal learning applications, the same data set is split and distributed amongst two or more operating environments. For example, active customer records may be contained within your customer’s CRM platform while inactive customer records housed in a cloud archive. In vertical learning applications, different types of data are distributed amongst two or more systems but share a common identifier which allows datasets to be created extracting data of interest. For example, patient clinical results such as blood pressure and heart rate may be contained within an EMR system while medical imagery for the same patient in a separate PACS system. Vertical learning enables the merging of the patient’s scans with clinical test results for dataset creation and modeling.
Federated Learning also provides your customers the ability to create larger supervised datasets with the ability to merge unlabeled data into a labeled dataset. This is a powerful capability that will result in higher predictive reliability as a function of achieving optimal modeling sample sizes.
The security framework for Analytics Anywhere was designed to meet the most stringent requirements of the US federal government intelligence agencies. Through our NIST 800-503 security compliance we are able to meet other industry standards such as PCI, HIPPA, and GDPR. Data level security is achieved using homomorphic encryption for data in transit, at rest and in memory and user access and activity logged on a continuous basis.