On-device personal AI requires models and agent software that can operate within fixed hardware limits. Developers also need ways to keep those systems improving after deployment.

Model builders are rethinking architectures designed around elastic cloud capacity. The edge offers fixed hardware but a far richer view of the user, which makes it the natural home for personal AI, according to Jeffrey Li (pictured), chief operating officer of Liquid AI Inc.

“The vision we have is that we should bring AI … closer to the user,” he said. “So we focus on building these AIs to run on the devices all around us.”

Li spoke with theCUBE Research’s Dave Vellante and John Furrier at the Fully Connected event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed on-device AI agents, fixed edge compute and observability loops. (* Disclosure below.)

Personal AI needs context within device limits

Liquid AI’s Liquid Context, which is optimized for Snapdragon processors, sits between models, agents and hardware. It uses device signals to build an understanding of who the user is and what they’re trying to accomplish, Li explained.

“What is the best form factor to capture that signal? It’s the devices in our pockets,” Li said. “It’s our phones, it’s our wearables, it’s our watches, it’s our PCs, it’s our cars.”

Agent harnesses, the software that turns models into functioning agents, also manage user context. Keeping that context in an ever-growing text file poses challenges on devices with limited resources, according to Li. Liquid AI uses its own models to decide which information to retain and how to compress it.

“The problem with devices is that you have fixed compute,” he said. “You have to fit within the zero-sum compute. That means a lot of the assumptions around how harnesses today are built no longer hold at the edge.”

The company’s work includes a collaboration with Mercedes-Benz Group AG to bring on-device AI to its cars. Keeping those agents aligned with user expectations long after deployment is the next step, Li noted.

“From here, we want to build these systems and these agents to be able to self-heal and improve and personalize on their own over time,” he said. “We’re building observability loops and continuous improvement loops that will improve both the model and the harness over time through natural usage.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Fully Connected event:

(* Disclosure: TheCUBE is a paid media partner for the Fully Connected event. Neither CoreWeave, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE