Huge advances in the branch of AI known as deep learning have been rocking everything from the board game of GO to self-driving cars to new drug discovery.
While cutting-edge research is exciting, businesses need practical deep learning tools that work seamlessly with their existing IT infrastructure to solve problems that impact their bottom line.
We’re partnering with NVIDIA to address this gap.
Python-focused deep learning frameworks such as TensorFlow, Caffe and Theano are popular in academia to solve unique research challenges. But they can take considerable effort to deploy to a business’s production stack because enterprise infrastructure is written in Java or, more broadly, for the Java virtual machine.
For many large companies, the infrastructure that powers their core business processes also determines which new technologies they adopt. So the tools they need are drastically different from those used in research.
With NVIDIA’s powerful DGX-1 AI supercomputer and Skymind’s open-source deep learning distribution, SKIL (Skymind Intelligence Layer), large organizations can leverage a single, unified platform that works natively with their existing software to build deep learning solutions.
A Package Deal
Running on the eight Tesla P100 SMX2 cards inside the NVIDIA DGX-1, SKIL creates a bridge between the dominant enterprise IT stacks that rely on Java and the hardware acceleration that is already widespread in academia and AI research.
Using the DGX-1’s massive parallel processing, SKIL makes it possible for organizations to train large neural networks quickly, while integrating other popular open-source tools in the Java ecosystem, such as the big-data storage engine Hadoop, distributed run-time Spark and message queue Kafka. These tools are central to big data management in enterprises, and deep learning needs big data to produce value.
As a result, enterprises can build bleeding-edge deep learning tools on top of their existing software using the DGX-1 and SKIL.
In popular media, self-driving cars and robot assistants are depicted as the future of AI. You may be surprised to learn that businesses use deep learning to solve problems closer to home.
The goal of any enterprise is to reduce costs, increase profit and comply with regulations. Here are some of the ways DGX-1 and SKIL can help businesses achieve these goals:
- Anomaly Detection – Payment fraud and network intrusion are multi-billion dollar problems faced by companies in industries as diverse as financial services and social networking. Deep learning can adapt to rapidly changing online behavior and stop scammers before revenue is lost or reputations are damaged.
- Recommender Systems – E-commerce giants like Amazon already use deep learning to create personalized recommendations for each buyer based on browsing and buying behavior. Research firm MarketingSherpa determined that these recommendations were responsible for 11.5% of all revenue.
- Resource Planning – Large manufacturers use deep learning to forecast demand and adjust production accordingly. Cloud and virtualization providers automate the management of resources to reduce operational costs.
- Predictive Trading – Financial institutions use deep learning to predict stock market movement and programmatically place trades and hedges accordingly.
These are just a few examples of how businesses are using AI to add value. The combination of massive GPU computing resources with a deep learning framework adapted to Java and the enterprise production stack makes the NVIDIA and Skymind offering a powerful tool for enterprise teams.