Three years ago, SAS boldly redefined model performance by revolutionizing how models operate closely with databases, in collaboration with SingleStore.
Since then, we have been innovating on how real-time data is brought to your models to deliver insights — without the need to move that data.
Today, the culmination of these efforts has been validated by OpenAI's acquisition of Rockset. SAS has achieved remarkable milestones, boosting model performance by 50% while nearly halving costs through their SAS with SingleStore solution.
Recognizing that serving real-time data to models is a pivotal bottleneck in model training and contextualization, SAS and SingleStore have pioneered an integrated solution. This innovation seamlessly fuses data infrastructure with model computation, optimizing the deployment of both data and models for unparalleled efficiency in generating outcomes and intelligence.
OpenAI’s strategic acquisition of Rockset underscores the critical role of real-time analytics databases. SAS foresaw this paradigm shift three years ago when they aligned with SingleStore — the worlds of AI models and databases have been one for a while now.
In OpenAI’s and Rockset’s announcement tweet regarding the acquisition, we see many of the design principles that are currently in production in SAS with SingleStore including storing data in a secure, high-performance framework, minimizing data movement and complexities at scale and providing easier access to all analytical data and models:
“We'll integrate Rockset’s technology across our products, empowering companies to transform their data into actionable intelligence.”
Venkat Venkataramani, CEO and co-founder of Rockset, emphasized:
“Rockset will become part of OpenAI, enhancing the retrieval infrastructure powering OpenAI’s suite of products. We will tackle the challenging database issues that AI applications encounter at massive scales.”
At SingleStore, we're already seeing use cases like this at play: A large logistics customer in Europe has made the move from running models on legacy data storage to SingleStore, finding overall performance of historical batch data supporting models now runs five times faster. Specifically, the read speed of one table from SingleStore versus historical storage for computation is twice as fast — and joins of data embedded in a model are 10-100 times quicker.
Looking ahead, the integration between SAS’s cutting-edge model platform and SingleStore’s real-time, high speed database promises an exciting future. This integration represents a pivotal moment — setting a new standard for enterprise ML, AI and intelligent, generative AI applications.
Learn more about SAS Viya — Data and AI Platform here.
