Uptime Industries is taking a wager on on-premise AI infrastructure with an “AI-in-a-box” handheld system referred to as Lemony AI.
The dimensions of a sandwich, Lemony AI can supposedly run a big language mannequin (LLM), AI brokers and AI workflows all on one node. Every Lemony requires solely 65 watts of energy to run, the equal of charging a laptop computer, and could be stacked and related to kind greater AI clusters.
Sascha Buehrle, co-founder and CEO of Uptime, advised TechCrunch that every node can maintain an LLM as much as 75 billion parameters in dimension, and may host open-source fashions or retrofitted variations of closed fashions. In a cluster of Lemony nodes, every system can run a unique mannequin.
Uptime is launching with partnerships with IBM and JetBrains to supply its clients simpler entry to AI fashions, together with IBM’s closed fashions.
Like most tech startups, the concept for Lemony AI stemmed from a aspect undertaking: Uptime’s co-founders, Buehrle and Ivan Kuleshov, had been making an attempt to determine if they might distribute language fashions on Raspberry Pi’s micro computer systems, and whereas this wasn’t initially associated to generative AI, when it labored, they began making an attempt to see what else these Raspberry Pis might deal with.
Once they realized operating fashions on native units might show the important thing to unlock extra AI adoption by organizations — say, enterprises that don’t wish to use cloud-based fashions — they began constructing a tool of their very own. They figured going small with an emphasis on information privateness might get firms to undertake AI faster.
“We have to construct one thing small which might go simply into the groups, and which [does] not require any organization-wide determination making — bringing on-premise generative AI options into the enterprise groups, principally,” Buehrle stated. Clusters of small and highly effective units may help the system develop in accordance with a buyer’s wants, he added.
The corporate says it has already seen robust demand from enterprise in heavily-regulated industries like finance, healthcare and legislation.
“Every part stays in your field,” Buehrle stated. “So your paperwork, your information, your emails — the fashions are hosted within the field, the brokers are operating regionally within the field, and nothing is leaving the field.”
Uptime has to this point raised $2 million in a seed funding spherical led by True Ventures, with participation from Alumni Ventures, JetBrains, and a few angel traders. The corporate will use the funds to additional develop its units.
The startup says the plan now could be to get the software program it has developed for its micro AI computer systems, Lemony OS, to work on different firms’ {hardware}, just like the Nvidia DGX Spark. It additionally needs to broaden the software program from its present, single-user focus to at least one that can be utilized by groups.
Lemony AI prices $499 a month to make use of, and could be accessed by as much as 5 customers.