Yutori, an AI assistant startup founded by former meta executives, raises $15 million to build an agent that goes beyond chat.

Yutori, a new AI startup from San Francisco, has come out of stealth mode, with a personal AI assistant who can actually complete a task with $15 million seed funds and a bold pitch: chatting about them.

Emerging from stealth, Yutori aims to build AI agents that complete trustworthy tasks with a “agent-first” approach that touches technology veterans from Meta, Google, Stanford and CMU.

The round was led by a radical venture, joined by Felicis and outstanding groups such as “AI Godmother” Fei-Fei Li, Jeff Dean, Elad Gil, Sarah Guo (convicted), and Sandhya Venkatachalam (Axiom). Other supporters include Replit CEO Amjad Masad, Concept COO Akshay Kothari, Vercel’s Guillermo Rauch, and several AI product leaders from Google and Meta.

Founded by former meta AI leaders Devi Parik, dhruv batra and abhishek das, yutori wants to take AI agents beyond the limits of the current chatbot experience. “Productivity isn’t about stuffing more on your day, it’s about getting your attention back on what’s really important,” Parif said. The team is building what they describe as “staff chiefs for everyone.” This is a personal AI assistant who can actually get things done online.

“We’re building the entire stack to be agent-first, from training our own models to generating product interfaces. Previously, Meta (Llama 3 & 4, Fair), Google (Gemini), Deepmind, Microsoft Research, Tesla Autopilot, Palantir, Minion, Riot Games, Headliner, 8tracks, Cmu, stanford, Georder, grainintir, tesla autopilot, palantir, minion, cmu, georford, startups said on their website.

This means not only answering questions and having conversations, but also handling suckling, food ordering, booking trips, dealing with transactions, booking deals, and handling reservations. The current generation of chatbots is often lacking when it comes to performing long, multi-step tasks, even with the latest leading language models (LLM). Yutori’s answer is to stop all centers around the LLMS and instead build around the agent itself.

This shift is what co-founder Abhishek Das calls the “agent-first” approach. “When we’re rethinking the interface between consumers and the web from scratch, we can’t focus solely on the AI ​​model, or orchestration or application layers,” he said. “To raise both the floor and ceiling of human productivity on the web, we need to innovate across the stack in a consistent, tightly coupled way.”

“When you’re rethinking the interface between consumers and the web from scratch – you
We focus solely on AI models or orchestration or application layers,” said co-founder Abhishek Das.
Innovate the entire stack in a consistent, combined way. That’s what we do at Yutori. ”

Yutori is working on a multi-agent system that allows tasks and subtasks to be executed in parallel. They use open source models and apply advanced post-training techniques to help agents adapt and recover when they are not going as planned. The web is unpredictable, and the agent that fails with the first hiccup is a non-starter for the user.

Chief scientist Dhruv Batra said: “It is the external environment that distinguishes chatbots from agents. The web is the ultimate digital environment. If you can do tasks digitally, you can do them over the web. yutori. “Yutori’s agent-first approach unlocks superhuman performance with this epic challenge.”

The founding team is nothing new to this space. Parif led Meta’s multimodal generation AI efforts, including projects such as Lama 3, Make Assen and EMU. Batra worked on a robotics demonstration embodied agents with Meta’s AR glasses and Boston dynamics. With his PhD research focused on vision language action agents, DAS received recognition for his work in building intelligent systems that allow him to see, speak and act. The rest of the team includes engineers and researchers from Meta, Google Gemini, CMU, Stanford, and other leading AI labs.

The new funds will be devoted to expanding the engineering and design teams and passing the first version of the product into the hands of users. The closed beta version is scheduled to be released this spring, and early adopters can sign up for yutori.com.

Yutori is not the only startup trying to build an AI agent, but it’s something to look at because it combines technical firepower, early support from top investors and clear product goals.

Yutori Team