Kinetica introduces the native Large Language Model

Embedded LLM enables organizations to operate SQL-GPT with improved privacy, security, and fine-tuning.

Kinetica, the speed layer for generative AI and real-time analytics, has introduced a native Large Language Model (LLM) that, when combined with Kinetica’s innovative architecture, enables users to conduct ad hoc data analysis using natural language on real-time, structured data. In contrast to public LLMs, no external API calls are necessary and data never exits the customer’s environment. This announcement follows Kinetica’s first OpenAI-integrated analytic database announcement.

In the aftermath of the excitement surrounding LLMs, businesses and government agencies are exploring novel ways to automate certain business tasks while ensuring the confidentiality of sensitive data that could be revealed through fine-tuning or prompt augmentation. Public LLMs, such as Open AI’s GPT 3.5, raise privacy and security concerns that can be mitigated by native offerings integrated into the Kinetica deployment and confined to the customer’s network perimeter.

Kinetica’s native LLM is not only more secure, but it is also suited to the syntax of the platform and the data definitions of specific industries, such as telecommunications, financial services, automotive, and logistics, amongst others. This results in more dependable and accurate SQL creation. This capability goes beyond that of normal SQL, and it ensures the effective handling of difficult jobs in order to improve decision-making for queries involving time series, graphs, and geographical relationships.

Kinetica’s method of fine-tuning places a higher priority on optimizing SQL production in order to guarantee accurate and consistent results. This contrasts with more conventional approaches, which place a higher priority on creativity and provide a variety of unique solutions. This ensures continued functionality for enterprises and consumers, providing them with peace of mind in regards to the outcomes of SQL queries.

According to Nima Negahban, Cofounder and CEO of Kinetica, Kinetica has led the market with its vectorized, real-time analytic database for processing sensor and machine data. Kinetica has been on the cutting edge of this technological development. With the integration of SQL-GPT, they extend this capacity to an altogether new horizon, which enables enterprises to unleash the actual potential of their real-time, structured data in a manner that has never been possible before.

Working together with Kinetica, the United States Air Force has been using sophisticated analytics to sensor data in order to rapidly recognize and respond to potential threats, thereby contributing to the maintenance of a safe and secure environment for all users of the national airspace system. The United States Air Force is currently employing Kinetica’s embedded LLM to identify anomalies and threats in our airspace using natural language.

President and Cofounder of Kinetica, Amit Vij, stated that the company believes in nurturing openness and embracing the diversity of generative AI models. Kinetica plans to bring out integration with other LLM platforms such as NVIDIA NeMo later this year for language to SQL when new state-of-the-art models become available.

The Kinetica database converts natural language queries to SQL and returns results in a matter of seconds, even for unknown or complex queries. In addition, Kinetica integrates multiple modes of analytics, such as time series, spatial, graph, and machine learning, which expands the range of queries that can be answered.