Samsung SDS, a global leader in digital transformation and innovation solutions, today announced the launch of Brightics AI Accelerator automated, distributed machine learning (AutoML) and deep learning (AutoDL) offerings. The solution is uniquely targeted to companies who seek to maximize their AI investment by improving AI productivity, reducing system and process inefficiencies, and driving unparalleled business value. With only a few lines of code activated through a single Jupyter Notebook or PyCharm IDE user interface, Brightics AI Accelerator enables professional machine learning scientists and engineers to save time and deliver value faster.
Today, model training time is a barrier to AI progress. Lacking insight into a feasible alternative to time-consuming migration of model training to distributed clusters, most AI teams simply accept the excessive time it takes to train a model. In fact, it takes up to 3 weeks to train an initial deep learning model and 3 to 9 months to complete a deep learning project end-to-end on a desktop. Brightics AI Accelerator eliminates these barriers by automating internal development workflows needed to set up, run and tear down or clean up the infrastructure that supports well-resourced, distributed machine or deep learning systems. Samsung SDS’ AutoML and AutoDL offerings increase AI team productivity up to 250x compared to traditional methods by shrinking the time it takes to deliver value from 2 months to an hour for machine learning projects and 9 months to only a couple weeks for deep learning projects. Companies that adopt Brightics AI Accelerator are also able to reduce cost by making more efficient use of IT, DevOps, machine learning, and deep learning resources.
“While companies across industries are continuing to invest in people, hardware, software, and services as they seek to capitalize on the promise and potential of AI, the majority are still struggling to realize this potential. We saw a great opportunity to develop a software platform that will equip data scientists to utilize AI resources to their full potential.” said Scott Koo, President of Samsung SDS America. “With the new AI Accelerator offerings, it is now possible to scale productivity seamlessly and efficiently while eliminating significant barriers to AI progress through a push-button experience. AI Accelerator editions are as easy to purchase, install and maintain as a consumer electronics product.”
Licensing Brightics AI Accelerator is an opportunity to lift and shift deep learning AI model training from desktop to automated, distributed clusters to scale data throughput near linearly with each additional GPU server up to 256 via high speed interconnect. The result of increasing data throughput nearly linearly is quick and efficient model convergence to state-of-the-art accuracy in supercomputer challenge speed. Since AI teams are able to run AutoML and AutoDL solutions fully on-premises or in their private cloud environment, proprietary data and models train inside the company’s secure data center or protected network. Additionally, Brightics AI Accelerator saves time and money spent on cloud resources by provisioning and orchestrating jobs and resource clean-ups, automatically.
Brightics AI Accelerator AutoDL software automates and accelerates deep learning model training using data-parallel, distributed synchronous Horovod Ring-All-Reduce Keras, TensorFlow, and PyTorch frameworks with minimal code. AutoDL exploits up to 512 GPUs per training job to produce a model in 1 hour versus 3 weeks using traditional methods. AutoDL eliminates the installation of any software or configuration per job and offers a painless experience in provisioning, running, monitoring and cleaning up jobs. For computer vision (images & video) and natural language processing (NLP) projects, AutoDL can shrink the entire project lifecycle from up to 9 months to only a couple weeks in a properly sized cluster using auto-training and grid search based hyper-parameter optimization.
Brightics AI Accelerator’s AutoML software automates and accelerates model training on tabular data by using automated model selection from Scikit-Learn, automated feature synthesis, and hyper-parameter search optimization. AutoML with synthetic feature generation exploits up to 256 CPU cores simultaneously to produce a Scikit-Learn model in 1 hour versus 2 months using manual feature engineering methods.
A Single-user edition is available for 15-day trial followed by purchase on the AWS Marketplace, and a Team edition is available for 15-day trial via xcelerator.ai.
About Samsung SDS America
Samsung SDS America (SDSA) is the U.S. subsidiary of Samsung SDS, a $9 billion global leader in digital transformation and innovation solutions. SDSA helps organizations optimize their productivity, make smarter business decisions, and improve their competitive positions in a hyper-connected economy using our enterprise software solutions for secure mobility, retail, DOOH, advanced analytics, automated machine learning, automated deep learning and contextual marketing.
Samsung SDS America