A brand new AI orchestration startup from the founders of Lithuanian unicorn Nord Safety is getting down to assist enterprises put their AI initiatives into manufacturing, with an preliminary give attention to bringing larger visibility, safety and adaptableness to giant language fashions (LLMs).
Nexos.ai, because the startup is named, is the handiwork of Tomas Okmanas (pictured above) and Eimantas Sabaliauskas, who constructed one of the vital recognizable manufacturers not solely in Lithuania, however in all of Europe. Nord Safety, finest identified for its flagship VPN product NordVPN, bootstrapped its method via its first 10 years earlier than succumbing to a bumper $100 million funding in 2022 at a $1.6 billion valuation (it later hit a $3 billion valuation throughout a subsequent fundraise).
Their new firm is exiting stealth at present with $8 million in funding from a slew of high-profile backers, together with lead investor Index Ventures, which has now made its first ever funding into Lithuania.
“We’ve identified of Tomas and the work that he’s finished for a few years, in order quickly as we heard that he was constructing a brand new firm within the AI house, and was lastly prepared to take enterprise capital cash at this [early] stage, we had been very keen,” Index Ventures’ accomplice Hannah Seal advised TechCrunch.
Different notable buyers embrace Creandum and Dig Ventures, and distinguished angels such because the CEOs of Datadog, Klarna, Supercell, and Wix additionally participated.
Capitalizing on a catalyst
At the moment, groups that need to put their AI into manufacturing have to attach myriad instruments, which doubtless entails recruiting and constructing groups with the required abilities. That is the place Nexos.ai needs to step in.
“I’ve seen that there’s a giant hole between working AI as pilots and going into manufacturing,” Okmanas advised TechCrunch in an interview. “Once you’re testing AI in your lab, it’d work and it may be helpful, however once you need to put it into manufacturing, particularly in enterprises, how do you guarantee excessive availability? How do you guarantee safety? How do you handle price?”
Nord Safety’s been round for greater than a decade, however 5 years in the past, it was folded into an umbrella firm known as Tesonet, an incubator with a portfolio of greater than two-dozen companies. One in every of these is web-hosting agency Hostinger, which lately added AI-enabled smarts to its web site constructing software. Okmanas, a Hostinger board member and shareholder, mentioned a number of the points they encountered served as a catalyst for what would finally turn out to be Nexos.ai.
“We wished to make use of AI in our web site builder, so we turned on OpenAI, we began testing it, and we put it in manufacturing,” Okmanas mentioned. “In August, we had $150,000 billed. For what? Why was it so costly? There was no visibility.”
And when OpenAI went down a handful of occasions, Okmanas was satisfied that one thing needed to be finished to make it simpler to deploy, handle and optimize the “more and more advanced ecosystem of AI fashions” that organizations might have.
Via a easy API (utility programming interface), clients can entry greater than 200 AI fashions, from big-name incumbents like OpenAI and Anthropic to smaller, area of interest LLMs. The concept is, if OpenAI goes down, an organization can briefly (and robotically) change to a special supplier with out breaking stride. Or if the prices concerned in accessing a particular LLM explode for no matter purpose, an organization can transition to a different one to maintain their prices down.
Nexos.ai additionally ushers “clever caching” into the combination — if a specific query is repeated by a number of customers, the system can flip to its personal database fairly than persevering with to have interaction the LLM, which might get costly.
On the safety and compliance fronts, Nexos.ai additionally prevents people from sending non-public information to LLM suppliers, or if an worker leaves an organization, their entry might be terminated instantly.
There’s no escaping the elephant within the room, although: One of many causes enterprises have been hesitant to embrace AI is the thorny subject of knowledge safety — healthcare corporations, banks, or insurance coverage companies can’t merely belief LLM suppliers with all their delicate info. It’s price noting that Hostinger itself was hit with an information breach in 2019 and NordVPN has additionally been hacked up to now — the kind of assaults that each one corporations face at present.
This raises questions round how Nexos.ai handles such information, provided that it’s internet hosting every little thing by itself infrastructure. Okmanas mentioned the corporate will doubtless supply self-hosting sooner or later, and that it already helps integrations with corporations’ personal inner LLMs.
It additionally has guardrails in place to detect when information, comparable to personally identifiable info (PII), is distributed to it — in such circumstances, it might probably re-route the info again to the originating firm’s personal LLMs or database. But when a question is generic, like a buyer asking an AI agent for particulars about their location and opening hours, then the question shall be dealt with on the Nexos.ai aspect.
From thought to inception
Going from an thought to formal incorporation took Nexos.ai round six weeks, and whereas the pace of securing the funding was largely right down to the founders’ pedigree, a giant a part of it was merely the timing.
“I really feel like we’ve lastly gone past the hype of AI, and now the real-world purposes are coming,” Seal added. “All the big enterprises are realizing that is actually significant, and they should undertake AI at scale. And now’s the time for the infrastructure to meet up with the fashions.”
The pace of execution, although, was substantively as a result of broader organizational setup at Tesonet, which has round 4,000 staff throughout its portfolio. This enabled Okmanas to shortly assemble a crew of round 30 individuals who he knew and trusted to work on Nexos.ai full-time.
“Now we have these groups that may actually be part of forces — they’ve been working collectively for therefore a few years, there’s no want to inform them what’s what,” Okmanas mentioned. “We’ll even be hiring from the skin, however that takes way more time.”
Nexos.ai’s platform is ready to launch by the top of March, although Okmanas mentioned it’s already working with a bunch of “beta clients and design companions.”