Nexos.ai, a new startup that helps enterprises run large language models in production, launched today with $8 million from investors.
Index Ventures led the funding round. It was joined by Creandum, Dig Ventures and the chief executives of several major tech firms. The list includes Datadog Inc., Wix.com Ltd., financial technology company Klarna Bank AB and Supercell Oy, a major video game developer.
Lithuania-based Nexos.ai was launched last year by Tomas Okmanas and Eimantas Sabaliauskas, the founders of cybersecurity provider Nord Security. The latter company develops a virtual private networking tool called NordVPN with more than 10 million users. It received a $3 billion valuation in its most recent funding round.
Nexos.ai is developing a platform that will make it easier for enterprises to build LLM-powered applications. The company says its software can restore workloads after an AI model outage, speed up troubleshooting and ease other maintenance tasks.
The platform is built around an application programming interface that provides access to more than 200 LLMs. The catalog includes both proprietary models and open-source algorithms such as Meta Platform Inc.’s Llama series. Because all the LLMs ingest prompts through the same API, developers can switch an application to a new model with less coding than would otherwise be required.
Nexos.ai is also capable of switching between LLMs automatically. If the model that an application relies on to process user queries goes offline, the platform reroutes prompts to another model. It can lower inference costs by sending simple requests to a smaller, less hardware-intensive LLM.
The company promises to cut hardware expenses further with a feature called semantic cache. It can store an LLM’s response to a query in a database and retrieve it the next time a user asks the same question. This removes the need for the AI model to generate the response from scratch, which avoids the associated costs.
The platform logs every prompt and LLM response in a searchable repository for troubleshooting purposes. Using the feature, developers can check how a model’s responses change over time to identify quality issues. Alongside individual prompts and the corresponding LLM responses, Nexos.ai also logs more than a dozen high-level metrics such as token usage.
Nexos.ai will provide guardrails for AI applications powered by its platform. According to the company, those guardrails ensure that prompts containing sensitive business data are sent to an enterprise’s internally-hosted LLMs rather than a third-party model. The platform also prevents models from generating harmful output.
The startup plans to launch its platform later this quarter. In the meantime, it’s testing the software with several enterprises that are applying it to use cases such as customer support automation.
Image: Nexos.ai
Your vote of support is important to us and it helps us keep the content FREE.
One click below supports our mission to provide free, deep, and relevant content.
Join our community on YouTube
Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.
THANK YOU