CIO Bulletin

SambaNova Systems – Leveraging the first generative AI platform specifically optimized for enterprises and government organizations

"Our purpose-built enterprise-scale AI platform is the technology backbone for the next generation of AI computing."

AI is changing the world and at SambaNova, the team believes that you don’t need unlimited resources to take advantage of the most advanced, valuable AI capabilities - capabilities that are helping organizations explore the universe, find cures for cancer, and giving companies access to insights that provide a competitive edge. The company has built an enterprise-ready AI platform from the ground up - intentionally designed for the most valuable and complex AI workloads of today and tomorrow. Using its platform to build a technology backbone for the next decade of AI innovation, organizations get pre-trained Foundation Models that truly transform the way they gain value from AI and deep learning. And, with its flagship offering, Dataflow-as-a-Service™, it helps them realize value 22x faster.

SambaNova was founded in 2017 in Palo Alto, California, by a group of industry luminaries, business leaders, and world-class innovators who understand AI. Today, the company has built an incredibly smart and motivated team dedicated to making a lasting impact on the industry and equipping its customers to thrive in the new era of AI.

SambaNova DataScale

SambaNova DataScale® is a fully integrated hardware-software system, powered by a dataflow architecture that enables organizations to train and deploy the most demanding deep learning, foundation model, and AI for Science workloads and achieve world record performance with the largest and most challenging models.

State-of-the-art deep learning, foundation models, and AI for Science is advancing quickly, defined by very large models and data, and very detailed sparse models. GPU-based infrastructure struggles to effectively manage these workloads without complex and costly overhead and compromises in performance and accuracy. DataScale’s dataflow architecture combined with large on-chip and system memory enable organizations to run workloads that cannot practically be handled by GPUs, including:

Large models, such as 100B+ parameter large language models
Large data, including true resolution 2D and 3D images
Detailed models, such as RNNs, sparse models, and other AI for Science workloads

As incremental performance gains from each generation of CPU and GPU declines, the demands of state-of-the-art AI and deep learning is increasing. Powered by the new Cardinal SN30™ RDU, DataScale SN30 delivers performance enhancements including world record GPT training that is 6x faster than DGX A100 systems (GPT 13B), 12.8x more memory capacity than DGX A100, and deployment automation and ease of scale capabilities. DataScale’s large on-chip and system memory enables it to handle the largest models and data with little to no parallelization required. The SambaFlow software manages scaling across any number of devices or configurations. Seamlessly scale up to 48 racks of DataScale systems with consistent rack-to-rack bandwidth and latency. This enables organizations to train and deploy the largest models without the cost, complexity, and overhead that GPUs require.

SambaNova Suite for generative AI

It is the most accurate generative AI models, optimized for enterprise and government organizations, deployed on-premises or in the cloud, and adapted with an organization’s data for greater accuracy. SambaNova is optimizing generative AI models for the enterprise with accuracy and is providing the ability for enterprises to adapt models with their own data and deploy anywhere – in the cloud or in a data center. Ultimately, this enables generative AI to evolve from merely hype and excitement to delivering real value for the enterprise. It Suite delivers a collection of the highest accuracy generative AI models, including both state-of-the-art open source models, as well as models that have been pre-trained by SambaNova including GPT and Bloom. It empowers customers to optimize these models with their own data to further increase accuracy, while also allowing them to retain ownership of models that have been adapted with their data.

SambaNova Suite has been developed as an open platform integrating innovations from ecosystem partners at every layer of the stack, including model development, data, and enterprise integration. SambaNova Suite is a full stack AI offering which can be deployed on-premises or in the cloud, so no data ever needs to leave the customer’s environment. Further, unlike consumer generative AI cloud offerings, SambaNova Suite is delivered on a dedicated model backbone for every customer.

Contact Center and Document Intelligence

Contact center business leaders are challenged to equally optimize for operational efficiency, customer satisfaction, employee retention, and risk and compliance, yet are forced to make tradeoffs as improvements to one call center metric results in decrements to the other KPIs. SambaNova Contact Center Intelligence enables enterprises to overcome the contact center productivity paradox by seamlessly integrating generative AI into existing contact center workflows, with no need to rip and replace existing tools. Receive detailed, real-time guidance to enable contact center employees to focus on the customer and guide them to deliver a better experience, without the need to worry about administrative overhead.

Powered by the most accurate generative AI, SambaNova Document Intelligence enables enterprise organizations to organize and extract intelligence from massive volumes of documents, to analyze information in long, complex documents, and to generate net new content based upon those documents. Use natural language to search massive volumes of documents for information and quickly get an easy to understand response.

Generative AI in Banking Organizations

Deep Learning has revolutionized what can be accomplished with Natural Language Processing (NLP) and large language models. Unlocking the potential of this technology requires more than simply training generic models with common terminology. It requires that the model be trained on the specific terminology of a given industry to ensure that the language and the context used is understood correctly. Only SambaNova empowers banking organizations to take advantage of the power of NLP with SambaNova Suite GPT BankingTM the first Deep Learning large language model purpose-built and pre-trained specifically for the financial services industry.

SambaNova Suite enables organizations to take advantage of the only fully integrated, full stack generative AI solution at scale, with the ability to be deployed in weeks rather than months. GPT Banking is an out-of-the box model, pre-trained and refined specifically for the needs of enterprise banking organizations. GPT Banking uniquely delivers unprecedented performance, accuracy, and time to value for the banking industry, far beyond what is possible with generic models.

Pre-trained for Banking Organizations

Models trained on generic language sets can miss critical distinctions that are specific to an industry. In common terms a check could mean an additional review to be sure something is correct, but for banks it is a method of payment. It is simple for a human to know the difference, but for an AI system, it can be much harder. That is why SambaNova pre-trained and then further refined the textual corpus used for GPT Banking, resulting in unprecedented accuracy for the financial services industry. The Banking solutions are delivered on the SambaNova Suite platform. Available as an on-premises implementation or delivered as a cloud service, SambaNova Suite eliminates the need to buy and install infrastructure, hire people to manage that infrastructure, and find data scientists to build and train models. With SambaNova Suite you can be up and running in a fraction of the time it would otherwise take, giving you rapid time to value while eliminating the uncertainty inherent in building your own models.

About | Rodrigo Liang

Rodrigo is a co-founder and the CEO of SambaNova Systems. Prior to joining SambaNova, Mr. Liang was responsible for SPARC Processor and ASIC Development at Oracle. He led the engineering organization responsible for the design of state-of-the-art processors and ASICs for Oracle’s enterprise servers.